Tex as A& M University
ECEN 314– Spring 2022
Signals and Systems
Module 4
Convolution
Learning Objectives
Be able to form the convolution of two signals.
Convolution
Convolution is an operation involving two signals. For two waveforms, 𝑥𝑥(𝑡𝑡) and
𝑦𝑦(𝑡𝑡), their convolution is given by
∞
𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 = � 𝑥𝑥 𝑢𝑢 𝑦𝑦 𝑡𝑡 − 𝑢𝑢 𝑑𝑑𝑑𝑑 .
−∞
Note that using a change of variables, 𝑣𝑣 = 𝑡𝑡 − 𝑢𝑢 this can be rewritten as
−∞
𝑧𝑧 𝑡𝑡 = − � 𝑥𝑥 𝑡𝑡 − 𝑣𝑣 𝑦𝑦 𝑣𝑣 𝑑𝑑𝑣𝑣
∞
∞
= � 𝑥𝑥 𝑡𝑡 − 𝑣𝑣 𝑦𝑦 𝑣𝑣 𝑑𝑑𝑑𝑑 .
−∞
Therefore 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 = 𝑦𝑦 𝑡𝑡 ∗ 𝑥𝑥 𝑡𝑡 .
In general, convolution is a rather nasty operation
and we often try to find various tricks to avoid
performing convolution, but sometimes we just
have no choice, so the next several slides will go
through numerous examples of computing
convolutions.
Example 1
𝑡𝑡−1 ∞
Suppose 𝑥𝑥 𝑡𝑡 = 𝑒𝑒 −𝑡𝑡 𝑢𝑢(𝑡𝑡) and 𝑦𝑦 𝑡𝑡 = rect . Find 𝑧𝑧 𝑡𝑡 = ∫−∞ 𝑥𝑥 𝑢𝑢 𝑦𝑦 𝑡𝑡 − 𝑢𝑢 𝑑𝑑𝑑𝑑
2
𝑥𝑥(𝑡𝑡) 𝑦𝑦(𝑡𝑡) 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
1 1
1
𝑡𝑡 𝑡𝑡 𝑢𝑢
2 𝑡𝑡 − 2 𝑡𝑡
Case I: 𝑡𝑡 < 0, 𝑥𝑥 𝑢𝑢 , 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
Since the two waveforms do not 1
overlap, the product 𝑥𝑥 𝑢𝑢 𝑦𝑦 𝑡𝑡 − 𝑢𝑢 = 0 𝑢𝑢
for all 𝑢𝑢 and therefore 𝑧𝑧 𝑡𝑡 = 0. 𝑡𝑡 − 2 𝑡𝑡
Case II: 0 < 𝑡𝑡 < 2, 𝑥𝑥 𝑢𝑢 , 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
𝑡𝑡 𝑡𝑡
𝑧𝑧 𝑡𝑡 = � 𝑒𝑒 −𝑢𝑢 𝑑𝑑𝑑𝑑 = −𝑢𝑢
− 𝑒𝑒 � = 1 − 𝑒𝑒 −𝑡𝑡 𝑢𝑢
0 0
𝑡𝑡 − 2 𝑡𝑡
Example 1 (continued)
Case III: 𝑡𝑡 > 2, 𝑥𝑥 𝑢𝑢 , 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
𝑡𝑡 𝑡𝑡
−𝑢𝑢 −𝑢𝑢
𝑧𝑧 𝑡𝑡 = � 𝑒𝑒 𝑑𝑑𝑑𝑑 = − 𝑒𝑒 � 𝑢𝑢
𝑡𝑡−2 𝑡𝑡−2
= 𝑒𝑒 − 𝑡𝑡−2
− 𝑒𝑒 −𝑡𝑡 = 𝑒𝑒 2 − 1 𝑒𝑒 , −𝑡𝑡 𝑡𝑡 − 2 𝑡𝑡
or equivalently,
𝑧𝑧 𝑡𝑡 = 1 − 𝑒𝑒 −2 𝑒𝑒 −(𝑡𝑡−2) .
Putting these three results together, we get:
0.9
0.8
0.7
0, 𝑡𝑡 < 0, 0.6
𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 = � 1 − 𝑒𝑒 −𝑡𝑡 , 0 < 𝑡𝑡 < 2, 0.5
𝑒𝑒 2 − 1 𝑒𝑒 −𝑡𝑡 , 𝑡𝑡 > 2.
z(t)
0.4
0.3
This procedure of breaking the convolution
0.2
integral into different cases depending on the
0.1
values of 𝑡𝑡 is pretty common.
0
-2 0 2 4 6 8
t
Example 2
Sometimes it is possible to evaluate a convolution geometrically and thereby avoid
∞
using calculus. Suppose 𝑥𝑥 𝑡𝑡 = 𝑦𝑦 𝑡𝑡 = rect 𝑡𝑡 . Find 𝑧𝑧 𝑡𝑡 = ∫−∞ 𝑥𝑥 𝑢𝑢 𝑦𝑦 𝑡𝑡 − 𝑢𝑢 𝑑𝑑𝑑𝑑
𝑥𝑥(𝑡𝑡) 𝑦𝑦(𝑡𝑡) 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
1 1
1
𝑡𝑡 𝑡𝑡 𝑢𝑢
−1/2 1/2 −1/2 1/2 𝑡𝑡 − 1/2 𝑡𝑡 + 1/2
𝑥𝑥 𝑢𝑢 , 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
Case I: 𝑡𝑡 < −1, 1
𝑧𝑧 𝑡𝑡 = 0. 𝑢𝑢
𝑡𝑡 − 1/2 𝑡𝑡 + 1/2 −1/2 1/2
𝑥𝑥 𝑢𝑢 , 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
Case II: −1 < 𝑡𝑡 < 0, 1
𝑡𝑡+1/2 𝑢𝑢
𝑧𝑧 𝑡𝑡 = ∫−1/2 𝑑𝑑𝑑𝑑 = t + 1.
−1/2 1/2
𝑡𝑡 − 1/2 𝑡𝑡 + 1/2
Example 2 (continued)
𝑥𝑥 𝑢𝑢 , 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
Case III: 0 < 𝑡𝑡 < 1, 1
1/2 𝑢𝑢
𝑧𝑧 𝑡𝑡 = ∫𝑡𝑡−1/2 𝑑𝑑𝑑𝑑 = −t + 1.
−1/2 1/2
𝑡𝑡 − 1/2 𝑡𝑡 + 1/2
𝑥𝑥 𝑢𝑢 , 𝑦𝑦(𝑡𝑡 − 𝑢𝑢)
Case IV: 𝑡𝑡 > 1, 1
𝑧𝑧 𝑡𝑡 = 0. 𝑢𝑢
−1/2 1/2 𝑡𝑡 − 1/2 𝑡𝑡 + 1/2
Putting these four results together, we get:
𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦(𝑡𝑡)
0, 𝑡𝑡 > 1, 1
𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 = �
1 − 𝑡𝑡 , 𝑡𝑡 < 1.
𝑡𝑡
Alternatively, 𝑧𝑧 𝑡𝑡 = tri(𝑡𝑡). −1 1
Example 3
Suppose 𝑥𝑥 𝑡𝑡 = 𝑦𝑦 𝑡𝑡 = 𝑒𝑒 −𝑡𝑡 𝑢𝑢(𝑡𝑡). Find the convolution 𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦(𝑡𝑡).
Example 4
Suppose 𝑥𝑥 𝑡𝑡 = 𝑒𝑒 −𝑡𝑡 𝑢𝑢(𝑡𝑡) and 𝑦𝑦 𝑡𝑡 = 𝑒𝑒 −2𝑡𝑡 𝑢𝑢(𝑡𝑡). Find the convolution 𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦(𝑡𝑡).
Example 5
Suppose 𝑥𝑥 𝑡𝑡 = sin(𝜔𝜔𝜔𝜔) and 𝑦𝑦 𝑡𝑡 = 𝑒𝑒 −2𝑡𝑡 𝑢𝑢(𝑡𝑡). Find the convolution 𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦(𝑡𝑡).
Properties of Convolution
Now that we have a handle on how to perform convolutions, let’s look at some
general properties of convolution and their implication on LTI systems.
1. Commutative Property - 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 = 𝑦𝑦 𝑡𝑡 ∗ 𝑥𝑥(𝑡𝑡).
When performing convolution, we can “flip + shift” either signal.
2. Associative Property – 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 ∗ 𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ [𝑦𝑦 𝑡𝑡 ∗ 𝑧𝑧 𝑡𝑡 ].
𝑥𝑥(𝑡𝑡) 𝑦𝑦(𝑡𝑡)
System 1 System 2
ℎ1 (𝑡𝑡) ℎ2 (𝑡𝑡)
These two systems
𝑥𝑥(𝑡𝑡) 𝑦𝑦(𝑡𝑡) are equivalent
Combined System
ℎ1 𝑡𝑡 ∗ ℎ2 (𝑡𝑡)
Properties of Convolution
3. Distributive Property - 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 + 𝑧𝑧 𝑡𝑡 = 𝑥𝑥 𝑡𝑡 ∗ 𝑦𝑦 𝑡𝑡 + 𝑥𝑥 𝑡𝑡 ∗ 𝑧𝑧(𝑡𝑡).
System 1
𝑥𝑥(𝑡𝑡) ℎ1 (𝑡𝑡) 𝑦𝑦(𝑡𝑡)
+
System 2
ℎ2 (𝑡𝑡) These two systems
are equivalent
𝑥𝑥(𝑡𝑡) 𝑦𝑦(𝑡𝑡)
Combined System
ℎ1 𝑡𝑡 + ℎ2 (𝑡𝑡)
4. Identity – 𝑥𝑥 𝑡𝑡 ∗ 𝛿𝛿 𝑡𝑡 = 𝛿𝛿 𝑡𝑡 ∗ 𝑥𝑥 𝑡𝑡 = 𝑥𝑥(𝑡𝑡)
𝛿𝛿(𝑡𝑡) System ℎ(𝑡𝑡) When it comes time to compute
ℎ(𝑡𝑡) convolutions, the delta function
is your new best friend.
ℎ 𝑡𝑡 = 𝛿𝛿 𝑡𝑡 ∗ ℎ(𝑡𝑡)
Example 6
𝑡𝑡−1
Suppose a certain LTI system has an impulse response of ℎ 𝑡𝑡 = rect . Find the
2
output of that system due to each of the following inputs:
1
(a) 𝑥𝑥1 𝑡𝑡 = rect(𝑡𝑡 − ),
2
(b) 𝑥𝑥2 𝑡𝑡 = δ(𝑡𝑡),
1
(c) 𝑥𝑥3 𝑡𝑡 = δ 𝑡𝑡 + rect(𝑡𝑡 − ).
2
