and Young (1983), and Oppenheim and Schafer (1989).

Continuous-Time and Discrete-Time Signals

In each of the above examples there is an input and an output, each of which is a time-varying
()
signal. We will treat a signal as a time-varying function, x t . For each time t, the signal has some
() ()
value x t , usually called x of t. Sometimes we will alternatively use x t to refer to the entire
signal x, thinking of t as a free variable.
() []
In practice, x t will usually be represented as a finite-length sequence of numbers, x n , in
which n can take integer values between 0 and N 1
, and where N is the length of the sequence.
[]
This discrete-time sequence is indexed by integers, so we take x n to mean the nth number in
sequence x, usually called x of n for short.
[] ()
The individual numbers in a sequence x n are called samples of the signal x t . The word
sample comes from the fact that the sequence is a discretely-sampled version of the continuous
signal. Imagine, for example, that you are measuring membrane potential (or just about anything
else, for that matter) as it varies over time. You will obtain a sequence of measurements sampled
at evenly spaced time intervals. Although the membrane potential varies continuously over time,
you will work just with the sequence of discrete-time measurements.
It is often mathematically convenient to work with continuous-time signals. But in practice,
you usually end up with discrete-time sequences because: (1) discrete-time samples are the only
things that can be measured and recorded when doing a real experiment; and (2) finite-length,
discrete-time sequences are the only things that can be stored and computed with computers.
In what follows, we will express most of the mathematics in the continuous-time domain. But
the examples will, by necessity, use discrete-time sequences.

Pulse and impulse signals. The unit impulse signal, written (t), is one at t = 0, and zero
everywhere else: (
(t) = 1 if t = 0
0 otherwise
The impulse signal will play a very important role in what follows.
One very useful way to think of the impulse signal is as a limiting case of the pulse signal,

(t): ( 1

(t) =
if 0 < t <

0 otherwise
The impulse signal is equal to the pulse signal when the pulse gets infinitely short:

(t) =
lim (t):

!0

2
 Unit step signal. The unit step signal, written u (t), is zero for all times less than zero, and 1
for all times greater than or equal to zero:
(
u(t) = 0 if t<0
1 if t0

P
Summation and integration. The Greek capital sigma, , is used as a shorthand notation
for adding up a set of numbers, typically having some variable take on a specified set of values.
Thus:
X5
i=1+2+3+4+5
i=1
P
The notation is particularly helpful in dealing with sums over discrete-time sequences:
3
X
x[n] = x[1] + x[2] + x[3]:
n=1

An integral is the limiting case of a summation:

Z1 1
X
x(t)dt =
lim
!0
x(k
)

t= 1 k= 1
For example, the step signal can be obtained as an integral of the impulse:
Zt
u(t) = (s)ds:
s= 1
0 ()
Up to s < the sum will be 0 since all the values of s for negative s are 0. At t the =0
cumulative sum jumps to 1 since (0) = 1
. And the cumulative sum stays at 1 for all values of t
()
greater than 0 since all the rest of the values of t are 0 again.
This is not a particularly impressive use of an integral, but it should help to remind you that it
is perfectly sensible to talk about infinite sums.

Arithmetic with signals. It is often useful to apply the ordinary operations of arithmetic to
=
signals. Thus we can write the product of signals x and y as z xy , meaning the signal made up
of the products of the corresponding elements:

z (t) = x(t) y (t)

Likewise the sum of signals x and y can be written z = +
x y . A signal x can be multiplied by
a scalar , meaning that each element of x is individually so multiplied. Finally, a signal may be
shifted by any amount:
z (t) = x(t s):

3
 = + + + +
... ...

Figure 1: Staircase approximation to a continuous-time signal.

Representing signals with impulses. Any signal can be expressed as a sum of scaled and
~( )
shifted unit impulses. We begin with the pulse or staircase approximation x t to a continuous
()
signal x t , as illustrated in Fig. 1. Conceptually, this is trivial: for each discrete sample of the
original signal, we make a pulse signal. Then we add up all these pulse signals to make up the
approximate signal. Each of these pulse signals can in turn be represented as a standard pulse
scaled by the appropriate value and shifted to the appropriate place. In mathematical notation:
1
X
x~(t) = x(k
)
(t k
)
:
k= 1


As we let approach zero, the approximation x ~(t) becomes better and better, and the in the limit
()
equals x t . Therefore,
1
X
x(t) =
lim
!0
x(k
)
(t k
)
:
k= 1
Also, as
! 0, the summation approaches an integral, and the pulse approaches the unit impulse:
Z1
x(t) = x(s) (t s) ds: (1)
1
In other words, we can represent any signal as an infinite sum of shifted and scaled unit impulses. A
digital compact disc, for example, stores whole complex pieces of music as lots of simple numbers
representing very short impulses, and then the CD player adds all the impulses back together one
after another to recreate the complex musical waveform.
This no doubt seems like a lot of trouble to go to, just to get back the same signal that we
originally started with, but in fact, we will very shortly be able to use Eq. 1 to perform a marvelous
trick.

Linear Systems

A system or transform maps an input signal x (t) into an output signal y(t):
y (t) = T [x(t)];
where T denotes the transform, a function from input signals to output signals.
Systems come in a wide variety of types. One important class is known as linear systems. To
see whether a system is linear, we need to test whether it obeys certain rules that all linear systems
obey. The two basic tests of linearity are homogeneity and additivity.

4
 Homogeneity. As we increase the strength of the input to a linear system, say we double it,
then we predict that the output function will also be doubled. For example, if the current injected
to a passive neural membrane is doubled, the resulting membrane potential fluctuations will double
as well. This is called the scalar rule or sometimes the homogeneity of linear systems.

Additivity. Suppose we we measure how the membrane potential fluctuates over time in
()
response to a complicated time-series of injected current x1 t . Next, we present a second (differ-
()
ent) complicated time-series x2 t . The second stimulus also generates fluctuations in the mem-
brane potential which we measure and write down. Then, we present the sum of the two currents
( )+ ( )
x1 t x2 t and see what happens. Since the system is linear, the measured membrane potential
fluctuations will be just the sum of the fluctuations to each of the two currents presented separately.

Superposition. Systems that satisfy both homogeneity and additivity are considered to be
linear systems. These two rules, taken together, are often referred to as the principle of superposi-
tion. Mathematically, the principle of superposition is expressed as:

T (x1 + x2 ) = T (x1 ) + T (x2 ) (2)

Homogeneity is a special case in which one of the signals is absent. Additivity is a special case in
which  = =1 .

Shift-invariance. Suppose that we inject a pulse of current and measure the membrane po-
tential fluctuations. Then we stimulate again with a similar pulse at a different point in time, and
again we measure the membrane potential fluctuations. If we havent damaged the membrane with
the first impulse then we should expect that the response to the second pulse will be the same as
the response to the first pulse. The only difference between them will be that the second pulse has
occurred later in time, that is, it is shifted in time. When the responses to the identical stimulus
presented shifted in time are the same, except for the corresponding shift in time, then we have
a special kind of linear system called a shift-invariant linear system. Just as not all systems are
linear, not all linear systems are shift-invariant.
In mathematical language, a system T is shift-invariant if and only if:

y (t) = T [x(t)] implies y (t s) = T [x(t s)] (3)

Convolution

Homogeneity, additivity, and shift invariance may, at first, sound a bit abstract but they are very
useful. To characterize a shift-invariant linear system, we need to measure only one thing: the way
the system responds to a unit impulse. This response is called the impulse response function of
the system. Once weve measured this function, we can (in principle) predict how the system will
respond to any other possible stimulus.

5
 Impulses
Impulse Impulse Response

Each Impulse Creates a

Scaled and Shifted Impulse Response

For
example

The sum of all the impulse responses

is the final system response

Figure 2: Characterizing a linear system using its impulse response.

The way we use the impulse response function is illustrated in Fig. 2. We conceive of the input
stimulus, in this case a sinusoid, as if it were the sum of a set of impulses (Eq. 1). We know the
responses we would get if each impulse was presented separately (i.e., scaled and shifted copies of
the impulse response). We simply add together all of the (scaled and shifted) impulse responses to
predict how the system will respond to the complete stimulus.
Now we will repeat all this in mathematical notation. Our goal is to show that the response (e.g.,
membrane potential fluctuation) of a shift-invariant linear system (e.g., passive neural membrane)
can be written as a sum of scaled and shifted copies of the systems impulse response function.

The convolution integral. Begin by using Eq. 1 to replace the input signal x (t) by its repre-
sentation in terms of impulses:
Z 1 
y (t) = T [x(t)] = T x(s) (t s) ds
2 1 3
1
X
= T
lim
4
!0 k = 1
x(k
)
(t k
)
5:

Using additivity,
1
X
y (t) =
lim
!0
T [x(k
)
(t k
)
]:
k= 1
Taking the limit, Z1
y (t) = T [x(s) (t s) ds]:
1

6
 past present future
0 0 0 1 0 0 0 0 0 input (impulse)

1/8 1/4 1/2 weights

0 0 0 1/2 1/4 1/8 0 0 0 output (impulse response)

0 0 0 1 1 1 1 1 1 input (step)

1/8 1/4 1/2 weights

0 0 0 1/2 3/4 7/8 7/8 7/8 7/8 output (step response)

Figure 3: Convolution as a series of weighted sums.

Using homogeneity, Z1
y (t) = x(s) T [ (t s)] ds:
1
()
Now let h t be the response of T to the unshifted unit impulse, i.e., h (t) = T [(t)]. Then by using
shift-invariance, Z 1
y (t) = x(s) h(t s) ds: (4)
1
Notice what this last equation means. For any shift-invariant linear system T , once we know its
()
impulse response h t (that is, its response to a unit impulse), we can forget about T entirely, and
()
just add up scaled and shifted copies of h t to calculate the response of T to any input whatsoever.
Thus any shift-invariant linear system is completely characterized by its impulse response h t . ()
The way of combining two signals specified by Eq. 4 is know as convolution. It is such a
widespread and useful formula that it has its own shorthand notation, . For any two signals x and
y , there will be another signal z obtained by convolving x with y ,
Z1
z (t) = x  y = x(s) y (t s) ds:
1

Convolution as a series of weighted sums. While superposition and convolution may sound
a little abstract, there is an equivalent statement that will make it concrete: a system is a shift-
invariant, linear system if and only if the responses are a weighted sum of the inputs. Figure 3
shows an example: the output at each point in time is computed simply as a weighted sum of the
inputs at recently past times. The choice of weighting function determines the behavior of the
system. Not surprisingly, the weighting function is very closely related to the impulse response of
the system. In particular, the impulse response and the weighting function are time-reversed copies
of one another, as demonstrated in the top part of the figure.

7
 Notes for Signals and Systems

5.1 DT LTI Systems and Convolution

Discrete-time systems that are linear and time invariant often are referred to as LTI systems. LTI
systems comprise a very important class of systems, and they can be described by a standard
mathematical formalism. To each LTI system there corresponds a signal h[n] such that the input-
output behavior of the system is described by

y[n] = x[k ]h[n k ]
k =
This expression is called the convolution sum representation for LTI systems. In addition, the
sifting property easily shows that h[n] is the response of the system to a unit-pulse input signal.
That is, for x[n] = [n],

y[n] = x[k ]h[n k ] = [k ]h[n k ] = h[n]
k = k =

Thus the input-output behavior of a discrete-time, linear, time-invariant system is completely

described by the unit-pulse response of the system. If h[n] is known, then the response to any
input can be computed from the convolution sum.

Derivation It is straightforward to show that a system described by the convolution sum, with
specified h[n], is a linear and time-invariant system. Linearity is clear, and to show time
invariance, consider a shifted input signal x[n] = x[n no ] . The system response to this input
signal is given by

y[n] = x[k ] h[n k ]
k =

= x[k no ] h[n k ]
k =
To rewrite this expression, change the summation index from k to l = k N, to obtain

y[n] = x[l ] h[n no l ]
l =
= y[n no ]
This establishes time invariance.

It is less straightforward to show that essentially any LTI system can be represented by the
convolution sum. But the convolution representation for linear, time-invariant systems can be
developed by adopting a particular representation for the input signal and then enforcing the
properties of linearity and time invariance on the corresponding response. The details are as
follows.

Often we will represent a given signal as a linear combination of basis signals that have certain
desirable properties for the purpose at hand. To develop a representation for discrete-time LTI
systems, it is convenient to represent the input signal as a linear combination of shifted unit pulse
signals: [n], [n 1], [n + 1], [n 2], . Indeed it is easy to verify the expression

50

x[n] = x[k ] [n k ]
k =
Here the coefficient of the signal [n k ] in the linear combination is the value x[k]. Thus, for
example, if n = 3, then the right side is evaluated by the sifting property to verify

x[k ] [3 k ] = x[3]
k =
We can use this signal representation to derive an LTI system representation as follows. The
response of an LTI system to a unit pulse input, x[n] = [n], is given the special notation y[n] =
h[n]. Then by time invariance, the response to a kshifted unit pulse, x[n] = [n k ] is
y[n] = h[n k ] . Furthermore, by linearity, the response to a linear combination of shifted unit
pulses is the linear combination of the responses to the shifted unit pulses. That is, the response to
x[n], as written above, is

y[n] = x[k ] h[n k ]
k =
Thus we have arrived at the convolution sum representation for LTI systems. The convolution
representation follows directly from linearity and time invariance no other properties of the
system are assumed (though there are some convergence issues that we have ignored). An
alternate expression for the convolution sum is obtained by changing the summation variable
from k to l = n k:

y[n] = h[l ]x[n l ]
k =
It is clear from the convolution representation that if we know the unit-pulse response of an LTI
system, then we can compute the response to any other input signal by evaluating the convolution
sum. Indeed, we specifically label LTI systems with the unit-pulse response in drawing block
diagrams, as shown below

The demonstration below can help with visualizing and understanding the convolution
representation.

Joy of Convolution (Discrete Time)

Response Computation

Evaluation of the convolution expression, given x[n] and h[n], is not as simple as might be
expected because it is actually a collection of summations, over the index k, that can take
different forms for different values of n. There are several strategies that can be used for
evaluation, and the main ones are reviewed below.

Analytic evaluation When x[n] and h[n] have simple, neat analytical expressions,
and the character of the summation doesnt change in complicated ways as n changes, sometimes
y[n] can be computed analytically.

51
Example Suppose the unit pulse response of an LTI system is a unit ramp,
h[n] = r[n] = n u[n]
To compute the response of this system to a unit-step input, write the convolution representation
as

y[n] = x[k ]h[n k ] = u[k ](n k ) u[n k ]
k = k =

= (n k ) u[n k ]
k =0
Note that in the second line the unit-step u[k] in the summand is removed, but the lower limit of
the sum is raised to zero, and this is valid regardless of the value of n. Now, if n < 0, then the
argument of the step function in the summand is negative for every k 0. Therefore y[n] = 0 for n
< 0. But, for n 0, we can remove the step u[n k ] from the summand if we lower the upper
limit to n. Then
n
y[n] = (n k ) = n + (n 1) + + 2 +1+ 0
k =0
Using the very old trick of pairing the n with the 0 , the (n 1) with the 1 , and so on, we see that
each pair sums to n. Counting the number of pairs for even n and for odd n gives

n(n + 1)
, n0
y[n] = 2
0, n<0
or, more compactly,
n(n + 1)
y[n] = u[n]
2

Remark Because of the prevalence of exponential signals, the following geometric-series

formulas for discrete-time signal calculations are useful for analytic evaluation of convolution.
For any complex number 1 ,
N 1 1 N
n =
n=0 1
For any complex number satisfying | |< 1 ,
1
n =
n=0 1

Graphical method This method is useful for more complicated cases. We

simply plot the two signals, x[k] and h[n k], in the summand versus k for the value of n of
interest, then perform lollypop-by-lollypop multiplication and plot the summand, and then add
up the lollypop values to obtain y[n].

Example To rework the previous example by the graphical method, writing

52

y[n] = x[k ]h[n k ]
k =
first plot h[k] as shown,

Then, for the value of n of interest, flip and shift. For n = 3, we plot h[3 k] and,
to facilitate the multiplication, plot x[k] immediately below:

Then lollypop-by-lollypop multiplication gives a plot of the summand,

Adding up the lollypop values gives

y[3] = 3 + 2 + 1 = 6
To compute, say, y[4], slide the plot of h[3 k] one sample to the right to obtain a plot of h[4 k]
and repeat the multiplication with x[k] and addition. In simple cases such as this, there is little
need to redraw because the pattern is clear. Even in complicated cases, it is often easy to identify
ranges of n where y[n] = 0, because the plots of x[k] and h[n k] are non-overlapping. In the
example, this clearly holds for n < 0.

LTI cleverness The third method makes use of the properties of linearity and
time invariance, and is well suited for the case where x[n] has only a few nonzero values. Indeed,
it is simply a specialization of the approach we took to the derivation of the convolution sum.

Example With an arbitrary h[n], suppose that the input signal comprises three nonzero lollypops,
and can be written as
x[n] = [n] + 2 [n 1] 3 [n 3]
Then linearity and time invariance dictate that
y[n] = h[n] + 2h[n 1] 3h[n 3]
Depending on the form of the unit-pulse response, this can be evaluated analytically, or by
graphical addition of plots of the unit-pulse response and its shifts and amplitude scales.

Remark The convolution operation can explode fail to be well defined for particular choices
of input signal and unit-pulse response. For example, with x[n] = h[n] = 1, for all n, there is no

53
value of n for which y[n] is defined, because the convolution sum is infinite for every n. In our
derivation, we did not worry about convergence of the summation. This again is a consequence of
our decision not to be precise about classes of allowable signals, or, more mathematically,
domains and ranges. The diligent student must always be on the look out for such anomalies.
Furthermore, there are LTI systems that cannot be described by a convolution sum, though these
are more in the nature of mathematical oddities than engineering legitimacies. In any case, this is
the reason we say that essentially any LTI system can be described by the convolution sum.

5.2 Properties of Convolution Interconnections of DT LTI Systems

The convolution of two signals yields a signal, and this obviously is a mathematical operation a
sort of weird multiplication of signals. This mathematical operation obeys certain algebraic
properties, and these properties can be interpreted as properties of systems and their
interconnections.

To simplify matters, we adopt a shorthand star notation for convolution and write

y[n] = ( x h)[n] = x[k ]h[n k ]
k =
Note that since, for any n, the value of y[n] in general depends on all values of the signals x[n]
and h[n], we use the more general operator notation style, In particular, we do not write
y[n] = x[n] h[n] because of the temptation to conclude that, for example, y[2] = x[2] h[2] .

Algebraic properties of the star operation are discussed below.

Commutativity Convolution is commutative. That is, ( x h)[n] = ( h x)[n] , or, in

complete detail,

x[k ]h[n k ] = h[k ]x[n k ] , for all n
k = k =
The proof of this involves the first standard rule for proofs in this course: Use change of variable
of summation. Beginning with the left side, replace the summation variable k by q = n k. Then
as k , q , but (unlike integration) it does not matter whether we sum from left-to-
right or right-to-left. Thus

( x h)[n] = x[k ]h[n k ] = x[n q ]h[q ]
k = q =

= h[q]x[n q] = (h x)[n]
q =

Using this result, there are two different ways to describe in words the role of the unit-pulse
response values in the input-output behavior of an LTI system. The value of h[n k] determines
how the nth value of the output signal depends on the k th value of the input signal. Or, the value
of h[q] determines how the value of y[n] depends on the value of x[n q].

Associativity Convolution is associative. That is,

( x (h1 h2 ))[n] = (( x h1 ) h2 )[n]

54
The proof of this property is a messy exercise in manipulating summations, and it is omitted.

Distributivity Convolution is distributive (with respect to addition). That is,

( x (h1 + h2 ))[ n] = ( x h1 )[ n] + ( x h2 )[n]

Of course, distributivity is a restatement of part of the linearity property of LTI systems and so no
proof is needed. The remaining part of the linearity condition is written in the new notation as
follows. For any constant b,
((bx) h)[n] = b ( x h)[n]

Shift Property This is simply a restatement of the time-invariance property, though the
notation makes it a bit awkward. For any integer no , if x[n] = x[n no ] , then
( x h ) [n] = ( x h ) [n no ]
Identity It is worth noting that the star operation has the unit pulse as an
identity element. Namely,
( x )[n] = x[n]
This can be interpreted in system-theoretic terms as the fact that the identity system, y[n] = x[n]
has the unit-pulse response h[n] = [n] . Also we can write ( ) [ n] = [ n] , an expression
that says nothing more than: The unit pulse is the unit-pulse response of the system whose unit-
pulse response is a unit pulse.

These algebraic properties of the mathematical operation of convolution lead directly to methods
for describing the input-output behavior of interconnections of LTI systems. Of course we use
block diagram representations to describe interconnections, but for LTI systems we label each
block with the corresponding unit-pulse response. For example,

Distributivity implies that the interconnection below

has the same input-output behavior as the system

55
Commutativity and associativity imply that the interconnections

both have the same input-output behavior as the system

Finally, we analyze the feedback connection

as follows, in an attempt to obtain a description for its input-output behavior. With the
intermediate signal e[n] labeled as shown, we can write
y[n] = ( g e)[n]
and
e[n] = x[n] (h y )[n]
as descriptions of the interconnection. Substituting the second into the first gives
y[n] = ( g x)[n] ( g h y )[n]
or, writing y[n] = ( y )[n] we get
( ( g h) y ) [n] = ( g x)[n]
However, we cannot solve for y[n] on the left side unless we know that the LTI system with
unit-pulse response ( g h)[n] is invertible. Lets stop here, and return to the problem of
describing the feedback connection after developing more tools.

But for systems without feedback, the algebraic rules for the convolution operation provide an
easy formalism for simplifying block diagrams. Typically it is easiest to start at the output signal
and write descriptions of the intermediate signals (labeled if needed) while working back toward
the input signal.

56
Example For the interconnected system shown below, there is no need to label internal signals as
the structure is reasonably transparent.

The output signal can be written as

y[n] = (h4 x)[n] + (h2 h1 x)[n] (h3 h1 x)[n]
= ( (h4 + h2 h1 h3 h1 ) x ) [n]
Thus the input-output behavior of the system is identical to the input-output behavior of the
system

5.3 DT LTI System Properties

Since the input-output behavior of a discrete-time LTI system is completely characterized by its
unit-pulse response, h[n], via the convolution expression

y[n] = x[k ]h[n k ] = h[k ]x[n k ]
k = k =
the input-output properties of the system can be characterized very precisely in terms of
properties of h[n].

Causal System An LTI system is causal if and only if h[n] = 0 for n < 0, that is,
if and only if h[n] is right sided.

The proof of this is quite easy from the convolution expression. If the unit-pulse response is right
sided, then the convolution expression simplifies to

y[n] = h[k ]x[n k ]
k =0
and, at any value of n, the value of y[n] depends only on the current and earlier values of the input
signal. If the unit-pulse response is not right sided, then it is easy to see that the value of y[n] at a
particular n depends on future values of the input signal.

57
 Memoryless System An LTI system is memoryless if and only if h[n] = c[n], for
some constant c. Again, a proof is quite easy to argue from the convolution expression.

Stable System An LTI system is (bounded-input, bounded-output) stable if and

only if the unit-pulse response is absolutely summable. That is,

| h[n] |
n =
is finite.

To prove this, suppose x[n] is a bounded input, that is, there is a constant M such that | x[n] | M
for all n. Then the absolute value of the output signal satisfies

| y[n] | = | h[k ]x[n k ] | | h[k ] || x[n k ] |
k = k =

M | h[k ] |
k =
Therefore, if the absolute summability condition holds, the output signal is bounded for any
bounded input signal, and we have shown that the system is stable.

To prove that stability of the system implies absolute summability requires considerable
cleverness. Consider the input x[n] defined by
1, h[n] 0
x[n] =
1, h[n] < 0
Clearly x[n] is a bounded input signal, and the corresponding response y[n] at n = 0 is

y[0] = h[k ] x[k ] = | h[k ] |
k = k =
Since the system is stable, y[n] is bounded, and therefore y[0] is bounded, and therefore the unit-
pulse response is absolutely summable.

Example The system with unit-pulse response

h[n] = (0.5)n u[n]
is a stable system since
1
| h[n] | = (0.5)n = =2
n = n=0 1 0.5
On the other hand, the system with unit-pulse response
h[n] = u[n 1]
is unstable.

Invertible System There is no simple characterization of invertibility in terms of

the the unit-pulse response. However, in particular examples it is sometimes possible to compute
the unit-pulse response of an inverse system, hI [ n] , from the requirement
(h hI )[n] = [n]
This condition expresses the natural requirement that a system in cascade with its inverse should
be the identity system.

58
Example To compute an inverse of the running summer, that is, the LTI system with unit pulse
response h[n] = u[n] , we must find hI [n] that satisfies

u[k ] hI [n k ] = [n]
k =
Simplifying the summation gives

hI [n k ] = [n]
k =0
It is clear that we should take hI [n] = 0 , for n < 0 . Using this result, for n = 0 the requirement
is

hI [k ] = hI [0] = 1
k =0
For n = 1 the requirement is

hI [1 k ] = hI [1] + hI [0] = 0
k =0
which gives hI [1] = 1 . Continuing for further values of n, it is clear that the inverse-system
requirement is satisfied by taking all remaining values of hI [n] to be zero. Thus the inverse
system has the unit pulse response
hI [n] = [n] [n 1]
Of course, it is easy to see that in general the output of this inverse system is the first difference of
the input signal.

5.4 Response to Singularity Signals

The response of a DT LTI system to the basic singularity signals is quite easy to compute. If the
system is described by

y[n] = x[k ]h[n k ]
k =
then the unit pulse response is simply y[n] = h[n] . If the input signal is a unit step, x[n] = u[n] ,
then

y[ n] = u[k ]h[n k ] = h[n k ]
k = k =0
n
= h[l ]
l =
In words, the unit-step response is the running sum of the unit-pulse response. Of course, if the
system is causal, that is, the unit-pulse response is right sided, then

59
Impulse Response

The impulse response of a linear system h (t) is the output of the system
at time t to an impulse at time . This can be written as

h = H( )

Care is required in interpreting this expression!

(t) h(t, 0)

0 t 0 t
H
(t ) h(t, )

0 t 0 t

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 3 / 55

Note: Be aware of potential confusion here:

When you write
h (t) = H( (t))
the variable t serves different roles on each side of the equation.

t on the left is a specific value for time, the time at which the output
is being sampled.
t on the right is varying over all real numbers, it is not the same t as
on the left.
The output at time specific time t on the left in general depends on
the input at all times t on the right (the entire input waveform).

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 4 / 55

 Assume the input impulse is at = 0,

h = h0 = H(0 ).

We want to know the impulse response at time t = 2. It doesnt

make any sense to set t = 2, and write

h(2) = H((2)) No!

First, (2) is something like zero, so H(0) would be zero. Second, the
value of h(2) depends on the entire input waveform, not just the
value at t = 2.

H 0
(t)
(2) h(t, 0) h(2, 0)
0 2 t 0 2 t

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 5 / 55

Time-invariance
If H is time invariant, delaying the input and output both by a time
should produce the same response

h (t) = h(t ).

In this case, we dont need to worry about h because it is just h shifted in

time.

(t) h(t)
0 t 0 t
H
(t )
h(t )
0 t 0 t

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 6 / 55

Linearity and Extended Linearity

Linearity: A system S is linear if it satisfies both

Homogeneity: If y = Sx, and a is a constant then

ay = S(ax).

Superposition: If y1 = Sx1 and y2 = Sx2 , then

y1 + y2 = S(x1 + x2 ).

Combined Homogeneity and Superposition:

If y1 = Sx1 and y2 = Sx2 , and a and b are constants,

ay1 + by2 = S(ax1 + bx2 )

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 7 / 55

Extended Linearity

Summation: If yn = Sxn for all n, an integer from ( < n < ),

and an are constants
!
X X
an yn = S an x n
n n

Summation and the system operator commute, and can be

interchanged.
Integration (Simple Example) : If y = Sx,
Z Z 
a( )y (t ) d = S a( )x(t )d

Integration and the system operator commute, and can be

interchanged.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 8 / 55

Output of an LTI System

We would like to determine an expression for the output y (t) of an linear

time invariant system, given an input x(t)

x y
H

We can write a signal x(t) as a sample of itself

Z
x(t) = x( ) (t) d

This means that x(t) can be written as a weighted integral of functions.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 9 / 55

Applying the system H to the input x(t),

y (t) = H (x(t))
Z 
= H x( ) (t)d

If the system obeys extended linearity we can interchange the order of the
system operator and the integration
Z
y (t) = x( )H ( (t)) d.

The impulse response is

h (t) = H( (t)).

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 10 / 55

Substituting for the impulse response gives
Z
y (t) = x( )h (t)d.

This is a superposition integral. The values of x( )h(t, )d are

superimposed (added up) for each input time .
If H is time invariant, this written more simply as
Z
y (t) = x( )h (t)d.

This is in the form of a convolution integral, which will be the subject of

the next class.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 11 / 55

Graphically, this can be represented as:

Input Output
(t) h(t)

0 t 0 t
(t ) h(t )
0 t 0 t
(x()d)(t ) (x()d)h(t )
x(t)

0 t 0 t
y(t)
x(t)

0 t 0 t
Z Z
x(t) = x()(t )d y(t) = x()h(t )d

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 12 / 55

System Equation
The System Equation relates the outputs of a system to its inputs.
Example from last time: the system described by the block diagram

x +
+
Z y
-

has a system equation

y 0 + ay = x.
In addition, the initial conditions must be given to uniquely specify a
solution.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 13 / 55

Solutions for the System Equation

Solving the system equation tells us the output for a given input.
The output consists of two components:

The zero-input response, which is what the system does with no input
at all. This is due to initial conditions, such as energy stored in
capacitors and inductors.

x(t) = 0 y(t)
0 t 0 t
H

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 14 / 55

 The zero-state response, which is the output of the system with all
initial conditions zero.

x(t) y(t)
0 t 0 t
H

If H is a linear system, its zero-input response is zero. Homogeneity

states if y = F (ax), then y = aF (x). If a = 0 then a zero input
requires a zero output.

x(t) = 0 y(t) = 0
0 t 0 t
H

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 15 / 55

Example: Solve for the voltage across the capacitor y (t) for an arbitrary
input voltage x(t), given an initial value y (0) = Y0 .

i(t) R
+
x(t) +

C y(t)

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 16 / 55

From Kirchhoffs voltage law

x(t) = Ri(t) + y (t)

Using i(t) = Cy 0 (t)

RCy 0 (t) + y (t) = x(t).
This is a first order LCCODE, which is linear with zero initial conditions.
First we solve for the homogeneous solution by setting the right side (the
input) to zero
RCy 0 (t) + y (t) = 0.
The solution to this is
y (t) = Ae t/RC
which can be verified by direct substitution.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 17 / 55

To solve for the total response, we let the undetermined coefficient be a

function of time
y (t) = A(t)e t/RC .
Substituting this into the differential equation
 
1
RC A0 (t)e t/RC A(t)e t/RC + A(t)e t/RC = x(t)
RC

Simplifying  
1 t/RC
A0 (t) = x(t) e
RC
which can be integrated from t = 0 to get
Z t  
1 /RC
A(t) = x( ) e d + A(0)
0 RC

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 18 / 55

Then

y (t) = A(t)e t/RC

Z t  
1 /RC
= e t/RC x( ) e d + A(0)e t/RC
0 RC
Z t  
1 (t )/RC
= x( ) e d + A(0)e t/RC
0 RC

At t = 0, y (0) = Y0 , so this gives A(0) = Y0

Z t  
1 (t )/RC
y (t) = x( ) e d + Y e t/RC .
RC | 0 {z }
|0 {z } zeroinput response
zerostate response

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 19 / 55

RC Circuit example
The impulse response of the RC circuit example is
1 t/RC
h(t) = e
RC
The response of this system to an input x(t) is then
Z t
y (t) = x( )h (t)d
Z0 t  
1 (t )/RC
= x( ) e d
0 RC

which is the zero state solution we found earlier.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 20 / 55

 Example:
High energy photon detectors can be modeled as having a simple
exponential decay impulse response.
0 Doshi et al.: LSO PET detector 1540

LE I. Summary results from the various lightguide configuration experi-

ts.
Photomultiplier
Energy Light Average
Scintillating
Number of
resolution collection peak-to- Crystal
crystals clearly
Coupler !FWHN %# efficiency !%# valley ratio resolved
Light Fibers
ect LSOa 13.0
ghtguidea 19.9 Light 100.0
40.6
10.0
2.5
9
7
V lens 27.2 28.0 2.5 7
era 35.0 12.6 6.0 6
Crystal
er taper 19.5 27.0 7.5 9

ergy resolution and light collection efficiency were measured with single
htguide elements.
ding the PMT socket containing the dynode resistor chain
s network, is 3 cm long, 3 cm wide, and 9.75 cm long.
These are used in Positiron Emmision
were defined. The detectors were then configured in coinci-
METHODSDETECTOR CHARACTERIZATION
Input is a sequence of impulses
Flood source histogram ping (photons).
a 1 mm diameter 22Na point source !same as used in
A detector module was uniformly irradiated with a 68Ge
nt source !2.6 " Ci#. The signals from the PS-PMT were
ated and digitized as described above in Sec. II D. The
wer energy threshold was set (Lecture
Cuff to approximately
3) $100 keV
h the aid of the threshold on the constant fraction dis-
minator and no upper energy threshold was applied.
Energy spectra Output is superposition of impulse responses (light).
Boundaries were drawn on the 2D position map to define
ook-up table !LUT# which relates position in the 2D his-
ram to the appropriate element in the LSO array. The raw
Input: Photons
mode data were then resorted and a histogram of total
se amplitudes !sum of the four position outputs# gener-
d for each crystal in the array. These energy spectra were
alyzed to determine the FWHM and the location of the
1 keV photopeak of each crystal. These two parameters
asure the energy resolution and light collection efficiency,
pectively.
Timing resolution
Two detectors were aligned facing each other, 15 cm
art, and connected in coincidence. A 22Na point source
8 " Ci# encapsulated in a 25.4 mm diameter, 3 mm thick
ar plastic disc with the activity in the central 1 mm was
ced in the center of the two detectors. For each detected
ncidence event, the sum of the four position signals for
h detector was sent to a constant fraction discriminator
ich generated timing pulses. The lower energy threshold
the CFD was set to approximately 100 keV. These two
ing pulses !one for each module# were in turn fed into a
ibrated time-to-amplitude converter !TAC# module. The
put from the TAC was then digitized to produce the tim-
spectrum.
Coincidence point spread function
Flood source histograms of both detectors were obtained
Cuff (Lecture 3)
Photon
From: Doshi et al, Med Phys. 27(7), p1535 July 2000
FIG. 5. A picture of the assembled detector module consisting of a 9!9
array of 3!3!20 mm3 LSO crystals coupled through a tapered optical fiber
bundle to a Hamamatsu R5900-C8 PS-PMT.
Tomography (PET) systems.
dence, 15 cm apart, and list-mode data was acquired by step-
Sec. III C# between the detectors in 0.254 mm steps. The
point source was scanned across the fifth row of the detector.
For each opposing crystal pair, the counts were recorded as a
function
ELE 301: of the point
Signals source position. A lower energyFall
and Systems win-
2011-12 21 / 55
dow of $100 keV was applied. The FWHM of the resulting
distribution for each crystal pair was determined to give the
intrinsic spatial resolution of the detectors.
E. Detector efficiency
A measure of the absolute detector efficiency was ob-
tained. A 18F point source with known activity !68 " Ci# was
Output: Light
placed 15 cm away from the face of the detector module. The
actual gamma-ray flux impinging on the detector face was
calculated from the solid angle subtended by the detector
module at the source. The constant fraction discriminator
was set to eliminate electronic noise ($100 keV# and the full
energy spectrum was obtained for each crystal over a fixed
time. A background measurement without the 18F point
t
source was also obtained to subtract the LSO background t
from the measurement. A lower energy window of 350 keV
was applied to all of the crystals and the number of counts
falling under the photopeak was calculated. The number of
counts detected was then divided by the total number of
gamma rays impinging on the detector module to obtain the
t
detector efficiency. t
IV. RESULTSDETECTOR CHARACTERIZATION
A. Flood source histogram results
An image of the flood histogram from one detector mod-
t
ule is shown in Fig. 6. All 81 crystals from the 9!9 LSO t
array are clearly visible. An average peak-to-valley ratio of
3.5 was obtained over the central row of nine crystals. Not
all crystals are uniformly spaced in the flood histogram after
applying Anger logic. This may be a result of the nonuni-
form tapering of the optical fiber taper, the nonuniform pack-
ing of the reflectance powder between the crystals, or most
ELE 301: Signals and Systems Fall 2011-12 22 / 55
Summary

For an input x(t), the output of an linear system is given by the

superposition integral
Z
y (t) = x( )h (t) d

If the system is also time invariant, the result is a convolution integral

Z
y (t) = x( )h(t ) d

The response of an LTI system is completely characterized by its

impulse response h(t).

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 23 / 55

Another expression for the superposition integral can be found by

substituting for = t 1 . Then d = d1 and 1 = t ,
Z
y (t) = x( )h(t )d

Z
= x(t 1 )h(t (t 1 ))d(1 )
Z
= x(t 1 )h(1 )d1 .

The block diagrams for a system using the impulse response:

x(t) y(t) x(t) y(t)

h(t) h(t)

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 24 / 55

Superposition Integral for Causal Systems

For a causal system h(t) = 0 for t < 0, and

Z
y (t) = x( )h (t) d.

Since h (t) = 0 for t < , we can replace the upper limit of the integral
by t Z t
y (t) = x( )h (t) d.

Only past and present values of x( ) contribute to y (t).

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 25 / 55

LTI System Response to a Sinusoidal Input

A LTI system has a real impulse response h(t). A sinusoidal input

x(t) = A cos(2f1 t + )

produces an output
Z
y (t) = h( ) [A cos(2f1 (t ) + )] d.

Using the identity cos(a b) = cos a cos b + sin a sin b,

Z
y (t) = A cos(2f1 t + ) h( ) cos(2f1 )d

Z
+A sin(2f1 t + ) h( ) sin(2f1 )d.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 26 / 55

Since h(t) is real,

y (t) = Hc (f1 )A cos(2f1 t + ) + Hs (f1 )A sin(2f1 t + ).

where
Z
Hc (f1 ) = h( ) cos(2f1 )d
Z

Hs (f1 ) = h( ) sin(2f1 )d

are real constants.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 27 / 55

We can then write the output as

y (t) = |H(f1 )|A cos (2f1 t + + H(f1 ))

(using the same trigonometric identity in reverse), where

q
|H(f )| = Hc2 (f1 ) + Hs2 (f1 )
H(f1 ) = tan1 (Hs (f1 )/Hc (f1 ))

Note that the response to a sinusoidal input is determined by a single

complex number H(f1 ), which determines the magnitude of the output,
and the phase shift.
A sinusoidal input is scaled and delayed by an LTI system, but is otherwise
unchanged.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 28 / 55

Summary

The response of an LTI system is completely characterized by its

impulse response h(t).
For an input x(t), the output of an linear system is given by the
superposition integral
Z
y (t) = x( )h (t) d

If the system is also time invariant, the result is a convolution integral

Z
y (t) = x( )h(t ) d

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 29 / 55

For a sinusoidal input at frequency f ,the output is

I a sinusoid at the same frequency,
I scaled in amplitude, and
I phase shifted.
This can be represented by a single complex number H(f ).

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 30 / 55

Convolution Evaluation and Properties

Review: response of an LTI system

Representation of convolution
Graphical interpretation
Examples
Properties of convolution

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 31 / 55

Convolution Integral

The convolution of an input signal x(t) with and impulse response h(t) is
Z
y (t) = x( )h(t ) d

= (x h)(t)

or
y = x h.
This is also often written as

y (t) = x(t) h(t)

which is potentially confusing, since the ts have different interpretations

on the left and right sides of the equation (your book does this).

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 32 / 55

Convolution Integral for Causal Systems

For a causal system h(t) = 0 for t < 0, and

Z
y (t) = x( )h(t ) d.

Since h(t ) = 0 for t < , the upper limit of the integral is t

Z t
y (t) = x( )h(t )d.

Only past and present values of x( ) contribute to y (t).

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 33 / 55

<t >t
Does not
x(t) y(t) contribute to y(t)

0 t 0 t
Z
! t
x(t) = x()(t )d y(t) = x( )h(t )d

If x(t) is also causal, x(t) = 0 for t < 0, and the integral further simplifies
Z t
y (t) = x( )h(t ) d.
0

Does not
contribute to y(t) <t >t
Does not
x(t) y(t) contribute to y(t)

0 t 0 t
! t ! t
x(t) = x( )(t )d y(t) = x( )h(t )d
0 0

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 34 / 55

Graphical Interpretation

An increment in input x( ) (t)d produces an impulse response

x( )h (t)d . The output is the integral of all of these responses
Z
y (t) = x( )h (t) d

Another perspective is just to look at the integral.

h (t) = h(t ) is the impulse response delayed to time

If we consider h(t ) to be a function of , then h(t ) is delayed
to time t, and reversed.
h(t ) h(t )

t t

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 35 / 55

This is multiplied point by point with the input,

x( )
h(t ) x( )h(t )

t t
Then integrate over to find y (t) for this t.

Graphically, to find y (t):

flip impulse response h( ) backwards in time (yields h( ))

drag to the right over t (yields h(( t)))
multiply pointwise by x (yields x( )h(t ))
Z
integrate over to get y (t) = x( )h(t ) d

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 36 / 55

Simple Example

2
x()
1

-1 0 1 2 3

1 h()

-1 0 1 2 3

2
h()
1

-1 0 1 2 3

2
h(t )
1

-1 0 1 2 3

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 37 / 55

2
x() x()
h(t ) t <0 h(t )
1 1 0<t <1

-1 0 1 2 3 -1 0 1 2 3

2
x() x()
h(t ) 1<t <2 2<t <3
1 1 h(t )

-1 0 1 2 3 -1 0 1 2 3

y(t) = (x h)(t)
2 2
x()
1 h(t ) t >3 1

-1 0 1 2 3 -1 0 1 2 3

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 38 / 55

Communication channel, e.g., twisted pair cable

x(t) y(t)
h(t)

Impulse response:
1.5

1 h(t)
h

0.5

0
0 2 4 6 8 10
tt

This is a delay 1, plus smoothing.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 39 / 55

Simple signaling at 0.5 bit/sec; Boolean signal 0, 1, 0, 1, 1, . . .

1
x(t)
0.5
u

0
0 2 4 6 8 10
tt
1
y(t)
0.5
y

0
0 2 4 6 8 10
tt

Output is delayed, smoothed version of input.

1s & 0s easily distinguished in y
Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 40 / 55
Simple signalling at 4 bit/sec; same Boolean signal

x(t)
1

0.5
u
0
0 2 4 6 8 10
t
1 y(t)

0.5
y

0
0 2 4 6 8 10
tt

Smoothing makes 1s & 0s very hard to distinguish in y .

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 41 / 55

Examples: Try these:

x(t) h(t) (x h)(t)

1 1 1

0 1 2 0 1 2 0 1 2

1 1 1

0 1 2 0 1 2 0 1 2

(t 1)
1 1 1

0 1 2 0 1 2 0 1 2

1 1 1

0 1 2 0 1 2 0 1 2

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 42 / 55

Properties of Convolution
For any two functions f and g the convolution is
Z
(f g )(t) = f ( )g (t ) d

If we make the substitution 1 = t , then = t 1 , and d = d1 .

Z
(f g )(t) = f (t 1 )g (1 ) (d1 )
Z
= g ( )f (t 1 ) d1

= (g f )(t)

This means that convolution is commutative.

Practically, If we have two signals to convolve, we can choose either to be
the signal we hold constant and the other to flip and drag.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 43 / 55

Simple Example (x*h)

2 2
x()
1 1 h()

-1 0 1 2 3 -1 0 1 2 3

xh 2 hx 2
x()
h(t ) x(t )
1 1
h()

-1 0 1 2 3 -1 0 1 2 3

y(t) = (x h)(t)
2

-1 0 1 2 3

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 44 / 55

Convolution is associative

If we convolve three functions f , g , and h

(f (g h))(t) = ((f g ) h)(t)

which means that convolution is associative.

Combining the commutative and associate properties,

f g h = f h g = = h g f

We can perform the convolutions in any order.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 45 / 55

Linearity
Convolution is also distributive,

f (g + h) = f g + f h

which is easily shown by writing out the convolution integral,

Z
(f (g + h))(t) = f ( ) [g (t ) + h(t )] d
Z Z
= f ( )g (t ) d + f ( )h(t ) d

= (f g )(t) + (f h)(t)

Together, the commutative, associative, and distributive properties mean

that there is an algebra of signals where

addition is like arithmetic or ordinary algebra, and

multiplication is replaced by convolution.
Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 46 / 55
Time-invariant

Convolution with a delayed signal gives a delayed output.

(f g )(t) = (f g )(t) = (f g )(t )

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 47 / 55

Properties of Convolution Systems

The properties of the convolution integral have important consequences
for systems described by convolution:

Convolution systems are linear: for all signals x1 , x2 and all , <,

h (x1 + x2 ) = (h x1 ) + (h x2 )

Convolution systems are time-invariant: if we shift the input signal x

by T , i.e., apply the input

x1 (t) = x(t T )

to the system, the output is

y1 (t) = y (t T ).

In other words: convolution systems commute with delay.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 48 / 55

 Composition of convolution systems corresponds to convolution of
impulse responses.
The cascade connection of two convolution systems y = (x f ) g
Composition

x w y
f g

is the same as a single system with an impulse response h = f g

x y
( f g)

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 49 / 55

Since convolution is commutative, the convolution systems are also

commutative. These two cascade connections have the same response

x w y
f g

x v y
g f

Many operations can be written as convolutions, and these all commute

(integration, differentiation, delay, ...)

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 50 / 55

Example: Measuring the impulse response of an LTI system.
We would like to measure the impulse response of an LTI system,
described by the impulse response h(t)
(t)
h(t)

0 t 0 t
h

This can be practically difficult because input amplitude is often limited. A

very short pulse then has very little energy.
A common alternative is to measure the step response s(t), the response
to a unit step input u(t)

s(t)
u(t)
0 t 0 t
h

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 51 / 55

The impulse response is determined by differentiating the step response,

s(t)
u(t) h(t)

0 t 0 t 0 t
d
h
dt

To show this, commute the convolution system and the differentiator to

produce a system with the same overall impulse response

(t)
u(t) h(t)
0 t 0 t 0 t
d
h
dt

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 52 / 55

Convolution Systems with Complex Exponential Inputs
If we have a convolution system with an impulse response h(t), and
and input e st where s = + j
Z
y (t) = h( )e s(t ) d

Z
= e st h( )e s d

We get the complex exponential back, with a complex constant

multiplier
Z
H(s) = h( )e s d

y (t) = e st H(s)

provided the integral converges.

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 53 / 55

H(s) is the transfer function of the system.

If the input is a complex sinusoid e jt ,
Z
H(j) = h( )e j d

y (t) = e jt H(j)

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 54 / 55

Summary

LTI systems can be represented as a the convolution of the input with

an impulse response.
Convolution has many useful properties (associative, commutative,
etc).
These carry over to LTI systems
I Composition of system blocks
I Order of system blocks
Useful both practically, and for understanding.
While convolution is conceptually simple, it can be practically difficult.
It can be tedious to convolve your way through a complex system.
There has to be a better way ...

Cuff (Lecture 3) ELE 301: Signals and Systems Fall 2011-12 55 / 55

 CONVOLUTION AND CORRELATION
http://www.tutorialspoint.com/signals_and_systems/convolution_and_correlation.htm Copyright tutorialspoint.com

Convolution
Convolution is a mathematical operation used to express the relation between input and output of
an LTI system. It relates input, output and impulse response of an LTI system as

y(t) = x(t) h(t)

Where y t = output of LTI

x t = input of LTI

h t = impulse response of LTI

There are two types of convolutions:

Continuous convolution

Discrete convolution

Continuous Convolution

y(t) = x(t) h(t)

= x()h(t )d
or

= x(t )h()d

Discrete Convolution

y(n) = x(n) h(n)

=
k=
x(k)h(n k)
or
=
k=
x(n k)h(k)
By using convolution we can find zero state response of the system.

Deconvolution
Deconvolution is reverse process to convolution widely used in signal and image processing.

Properties of Convolution

Commutative Property
 x1 (t) x2 (t) = x2 (t) x1 (t)

Distributive Property

x1 (t) [x2 (t) + x3 (t)] = [x1 (t) x2 (t)] + [x1 (t) x3 (t)]

Associative Property

x1 (t) [x2 (t) x3 (t)] = [x1 (t) x2 (t)] x3 (t)

Shifting Property

x1 (t) x2 (t) = y(t)

x1 (t) x2 (t t0 ) = y(t t0 )
x1 (t t0 ) x2 (t) = y(t t0 )
x1 (t t0 ) x2 (t t1 ) = y(t t0 t1 )

Convolution with Impulse

x1 (t) (t) = x(t)

x1 (t) (t t0 ) = x(t t0 )

Convolution of Unit Steps

u(t) u(t) = r(t)

u(t T1 ) u(t T2 ) = r(t T1 T2 )
u(n) u(n) = [n + 1]u(n)

Scaling Property

If x(t) h(t) = y(t)

1
then x(at) h(at) = y(at)
|a|

Differentiation of Output

if y(t) = x(t) h(t)

dy(t) dx(t)
then
dt
= dt
h(t)
or

dy(t) dh(t)
dt
= x(t) dt

Note:

Convolution of two causal sequences is causal.

Convolution of two anti causal sequences is anti causal.

Convolution of two unequal length rectangles results a trapezium.

Convolution of two equal length rectangles results a triangle.

 A function convoluted itself is equal to integration of that function.

Example: You know that u(t) u(t) = r(t)

According to above note, u(t) u(t) = u(t)dt = 1dt = t = r(t)
Here, you get the result just by integrating u(t) .

Limits of Convoluted Signal

If two signals are convoluted then the resulting convoluted signal has following range:

Sum of lower limits < t < sum of upper limits

Ex: find the range of convolution of signals given below

Here, we have two rectangles of unequal length to convolute, which results a trapezium.

The range of convoluted signal is:

Sum of lower limits < t < sum of upper limits

1 + 2 < t < 2 + 2
3 < t < 4
Hence the result is trapezium with period 7.

Area of Convoluted Signal

The area under convoluted signal is given by Ay = Ax Ah
Where A x = area under input signal

A h = area under impulse response

A y = area under output signal

Proof: y(t) = x()h(t )d
Take integration on both sides

y(t)dt = x()h(t )ddt

= x()d h(t )dt
We know that area of any signal is the integration of that signal itself.

Ay = Ax Ah
DC Component
DC component of any signal is given by
area of the signal
DC component = period of the signal

Ex: what is the dc component of the resultant convoluted signal given below?

Here area of x1 t = length breadth = 1 3 = 3

area of x2 t = length breadth = 1 4 = 4

area of convoluted signal = area of x1 t area of x2 t

= 3 4 = 12

Duration of the convoluted signal = sum of lower limits < t < sum of upper limits

= -1 + -2 < t < 2+2

= -3 < t < 4

Period=7
area of the signal
Dc component of the convoluted signal = period of the signal

Dc component = 12
7

Discrete Convolution
Let us see how to calculate discrete convolution:

i. To calculate discrete linear convolution:

Convolute two sequences x[n] = {a,b,c} & h[n] = [e,f,g]

Convoluted output = [ ea, eb+fa, ec+fb+ga, fc+gb, gc]

Note: if any two sequences have m, n number of samples respectively, then the resulting
convoluted sequence will have [m+n-1] samples.

Example: Convolute two sequences x[n] = {1,2,3} & h[n] = {-1,2,2}

Convoluted output y[n] = [ -1, -2+2, -3+4+2, 6+4, 6]

= [-1, 0, 3, 10, 6]

Here x[n] contains 3 samples and h[n] is also having 3 samples so the resulting sequence having
3+3-1 = 5 samples.

ii. To calculate periodic or circular convolution:

Periodic convolution is valid for discrete Fourier transform. To calculate periodic convolution all
the samples must be real. Periodic or circular convolution is also called as fast convolution.

If two sequences of length m, n respectively are convoluted using circular convolution then
resulting sequence having max [m,n] samples.

Ex: convolute two sequences x[n] = {1,2,3} & h[n] = {-1,2,2} using circular convolution

Normal Convoluted output y[n] = [ -1, -2+2, -3+4+2, 6+4, 6].

= [-1, 0, 3, 10, 6]

Here x[n] contains 3 samples and h[n] also has 3 samples. Hence the resulting sequence obtained
by circular convolution must have max[3,3]= 3 samples.

Now to get periodic convolution result, 1st 3 samples [as the period is 3] of normal convolution is
same next two samples are added to 1st samples as shown below:

Circular convolution result y[n] = [9 6 3]

Correlation
Correlation is a measure of similarity between two signals. The general formula for correlation is



x1 (t)x2 (t )dt

There are two types of correlation:

Auto correlation

Cros correlation

Auto Correlation Function

It is defined as correlation of a signal with itself. Auto correlation function is a measure of similarity
between a signal & its time delayed version. It is represented with R$$ .

Consider a signals xt . The auto correlation function of xt with its time delayed version is given by

R11 () = R() = x(t)x(t )dt [+ve shift]


= x(t)x(t + )dt [-ve shift]

Where = searching or scanning or delay parameter.

If the signal is complex then auto correlation function is given by

R11 () = R() = x(t)x (t )dt [+ve shift]


= x(t + )x (t)dt [-ve shift]

Properties of Auto-correlation Function of Energy Signal

Auto correlation exhibits conjugate symmetry i.e. R $$ = R*$$
Auto correlation function of energy signal at origin i.e. at =0 is equal to total energy of that
signal, which is given as:

R 0 = E = | x(t) |2 dt

Auto correlation function 1 ,

Auto correlation function is maximum at =0 i.e |R $$ | R 0

Auto correlation function and energy spectral densities are Fourier transform pairs. i.e.

F. T [R()] = ()

() = R()ej d
R() = x() x()

Auto Correlation Function of Power Signals

The auto correlation function of periodic power signal with period T is given by
T
1

2
R() = lim x(t)x (t )dt
T T T
2
Properties

Auto correlation of power signal exhibits conjugate symmetry i.e. R() = R ()

Auto correlation function of power signal at = 0 atorigin is equal to total power of that
signal. i.e.

R(0) =
Auto correlation function of power signal 1 ,

Auto correlation function of power signal is maximum at = 0 i.e.,

|R()| R(0)
Auto correlation function and power spectral densities are Fourier transform pairs. i.e.,

F. T [R()] = s()

s() = R()ej d
R() = x() x()
Density Spectrum
Let us see density spectrums:

Energy Density Spectrum

Energy density spectrum can be calculated using the formula:

E= | x(f) |2 df

Power Density Spectrum

Power density spectrum can be calculated by using the formula:
2
P =
n= | Cn |

Cross Correlation Function

Cross correlation is the measure of similarity between two different signals.

Consider two signals x1 t and x2 t . The cross correlation of these two signals R 12 () is given by

R12 () = x1 (t)x2 (t ) dt [+ve shift]


= x1 (t + )x2 (t) dt [-ve shift]

If signals are complex then

R12 () = x1 (t)x2 (t ) dt [+ve shift]


= x1 (t + )x2 (t) dt [-ve shift]


R21 () = x2 (t)x1 (t ) dt [+ve shift]


= x2 (t + )x1 (t) dt [-ve shift]

Properties of Cross Correlation Function of Energy and Power Signals

Auto correlation exhibits conjugate symmetry i.e. R 12 () = R21 () .

Cross correlation is not commutative like convolution i.e.

R12 () R21 ()

If R12 0 = 0 means, if x1 (t)x2 (t)dt = 0 , then the two signals are said to be orthogonal.

T
For power signal if limT 1 2
T x(t)x (t) dt then two signals are said to be orthogonal.
T
2

Cross correlation function corresponds to the multiplication of spectrums of one signal to the
complex conjugate of spectrum of another signal. i.e.

R12 () X1 ()X2 ()
This also called as correlation theorem.

Parsvel's Theorem
Parsvel's theorem for energy signals states that the total energy in a signal can be obtained by the
spectrum of the signal as

E= 1
2
|X()|2 d
Note: If a signal has energy E then time scaled version of that signal xat has energy E/a.