Scribd
Upload
10 views

Lecture 13: Causality of LSI Systems: Prof. Vikram M Gadre Indian Institute of Technology, Bombay January 15, 2015

Uploaded by

Arunmozhli
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
10 views

Lecture 13: Causality of LSI Systems: Prof. Vikram M Gadre Indian Institute of Technology, Bombay January 15, 2015

Uploaded by

Arunmozhli
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Lecture 13: Causality of LSI systems

Prof. Vikram M Gadre


Indian Institute of Technology, Bombay
January 15, 2015

IIT Bombay, 2015


This work is licensed under a Creative Commons
Attribution-ShareAlike 4.0 International License.
IITBombayX EE210.1X

1 Associativity and Commutativity of LSI Sys-


tems
Two more properties of system, namely associativity and commutativity are
explored here. These properties are not "new" in the sense that some basic
operations like addition and multiplication follow these properties. Note: Both
discrete and continuous systems are considered together in the following discus-
sion.

1.1 Associativity

In the above image, we input signal x, either x(t) for continuous system or
x[n] for discrete one, to LSI system S1 ( having its impulse response h1 (t) or
h1 [n]), which can be considered continuous or discrete, depending on the case.
The output of system S1 is fed to LSI system S2 ( having its impulse response
h2 (t) or h2 [n]), which also can be continuous or discrete. The output of S2
is y, either y(t) or y[n] in continuous and discrete case respectively. That is,
the whole setup can be considered either of the continuous or the discrete case.
In the continuous case, both the systems as well as the relevant signals are
continuous, whereas in discrete case they all operate on a discrete independent
variable.
Note here, that the output obtained from the first system is (x ∗ h1 ), continuous
or discrete. As this signal is fed to the S2 , we get output as

y = (x ∗ h1 ) ∗ h2

Now Associativity talks about the equivalence of systems put in series or


cascade. Mathematically,

(x ∗ h1 ) ∗ h2 = x ∗ (h1 ∗ h2 )

Now if we have a LSI system S3 , whose impulse response is h1 ∗h2 , then from
above equation it is clear that it will act exactly like these two systems S1 and
S2 would have acted when applied in series. This is the physical significance of
the associativity.

So in short, if we put two systems in a cascade, we can replace them with one
single system whose impulse response is equal to the convolution of the impulse
responses of these two systems.

2
IITBombayX EE210.1X

1.2 Commutativity
Let us apply signal x to system S1 and then its output to system S2 , both LSI,
to get output as y. Commutativity states that the order in which these systems
come is irrelevant to the result i.e. we will get the same output in both the cases.
It states that regardless of which of the two systems comes first, the output y
is going to be the same.

So the equivalent system which we talked about in the Associativity, is un-


affected if the order of these two systems is reversed.

We can also generalize this to n cascaded LSI systems and say that for any
permutation of the given n systems, the output is going to be the same. The
proof of this is a bit more intricate and is left an as exercise to the student. The
proof can be done by using the method of mathematical induction.

1.3 What if my systems are not LSI?


All of the above discussion is based on the assumption that all the systems we
are considering are LSI in nature. While it may be tempting to generalize these
seemingly simple results to all systems, and many times we tend to forget to
confirm that the system in question is LSI or not before applying such proper-
ties, we must be careful.
It turns out that the commutativity and associativity may not hold if the sys-
tems S1 and S2 are not both LSI. For example, Let’s say system S1 is given
by the equation y(t) = (x(t))2 and S2 by y(t) = x(t) + 1. We have taken
here continuous variable signals and systems, but the same discussion is valid
for the discrete case as well. So if we apply first S1 and then S2 , we will get
y(t) = (x(t))2 + 1 as the output, while reversing the order, i.e. first applying
S2 and then S1 would yield y(t) = (x(t) + 1)2 . So the both outputs are clearly
different.
But this does not mean that we can hurry and conclude that the resultant sys-
tems will never be the same. Consider, say, S1 as y(t) = x(t) + 4 and S2 as
y(t) = x(t) − 4. Now in this case, regardless of which system comes first, we get
output y(t) = x(t).
Basically, if even one of the systems is not LSI, we cannot say anything about
whether those systems follow associativity or commutativity beforehand: we
must examine those systems carefully and determine the answer for the individ-
ual case.

3
IITBombayX EE210.1X

2 Causality of a Linear Shift Invariant System


from its Impulse Response
We have noted before that a LSI system can be represented by its impulse
response, i.e. we know everything about a system once we know its impulse
response. We have seen system properties like causality and stability: according
to the previous statement, we should be able to know whether a system is
stable and/or causal just by looking at and examining its impulse response.
The conditions turn out to be very similar for continuous and discrete variable
systems and their respective impulse responses.

2.1 Causality of continuous variable LSI systems


We say that a system is causal when its output depends only on the past and the
current inputs, and not on the future inputs. We know that, for a continuous
independent variable system S‘ having impulse response ht, we can relate its
input and output by the equation
Z +∞
y(t) = h(τ )x(t − τ ) dτ
−∞

Note that as τ goes from −∞ to +∞, t − τ also goes from −∞ to +∞.


Note that τ being positive corresponds to the past ( as t − τ < t when τ is
positive and ’t’ is the current timeframe) while τ negative to the future and 0
to the current time. Now for the system being causal, we want to eliminate the
future inputs, so we must have h(τ ) 0 when τ is negative, so that there’s no
contribution of the future terms to the y(t), the output. Note that h(τ ) = 0
when τ < 0 is sufficient condition for causality of the system. It is easy to prove
that this condition is also necessary for the system to be causal i.e. there can be
no system which is causal that violates this condition. The proof can be done
by using the method of proof by contradiction.

2.2 Causality of discrete variable LSI systems


For discrete systems, we have the following equation relating output to the
input(s):
+∞
X
y[n] = h[κ]x[n − κ]
κ=−∞

Similar to the argument for the continuous case, κ < 0 corresponds to the future,
κ > 0 to the past and κ = 0 to the present/current input. So causality will be
ensured if all the future inputs are cut off, i.e. h[κ] = 0, ∀κ < 0. This is the
sufficient condition for system to be causal. We can also prove the sufficiency
by one more way:
Consider signals x1 [n] and x2 [n] such that x1 [n] = x2 [n] ∀n ≤ n0 .
Note that x1 [n] may not be the same as x2 [n] for any n ≥ n0 . Also, from given
condition( we are assuming the condition we arrived at to reach at the conclusion
that the system is causal) h[κ] = 0∀κ < 0
Now if the system is causal, we must be able to reach y1 [n] = y2 [n] ∀n ≤ n0 .

4
IITBombayX EE210.1X

Consider
+∞
X
y1 [n] = h[κ]x1 [n − κ], n ≤ n0
κ=0

Note that we started the sum from 0 as before zero, in−∞ to 0 region, we
already know that h[κ] = 0. Now observe the summation. it contains x1 [n − κ].
But as κ is less than 0, we have n − κ > n, or x1 [n − κ] = x2 [n − κ].
So the sum can be converted to
+∞
X
y1 [n] = h[κ]x2 [n − κ]
κ=0

And the right hand side of the equation is essentially y2 [n].


Hence y1 [n] = y2 [n]∀n ≤ n0 . Hence the sufficiency of the condition is proved.
The necessity can also be proved, by using proof by contradiction.

You might also like