Double CLICK on Right Brackets to expand the Sections

1.Signals

1.2 Elementary Signals

Whenever we try to attach numerical quantities to a phenomenon, we need a reference frame. This starts
from the moment we learn how to count, when our fingers become our own reference frame, to the time
we learn how to write numbers when the powers of ten become the reference frame. The worse which
can happen to a person is to find him/herself into a situation where there is no reference frame: in the
middle of the desert or of the ocean without instruments, or among people speaking a totally foreign
language. In other words without a reference frame we are lost!
In signal processing we have the same problem: we need an appropriate reference frame so that we can
place the signals we try to analyze. Since we deal with signals, any reference frame of interest has to be
made of signals with well know properties, which reflect the kind of information we want to extract. For
example if we hear a musical note, and we want to determine which note it is in the scale, we need to
compare it with sinusoidal signals (ie "tones") at all frequencies and determine which frequency is the
closest.
In what follows we present some of the elementary discrete time signals which will be at the basis of the
analysis developed in the rest of the book.

1.2.1 Discrete Time Unit Impulse

The unit impulse d@nD is the most "elementary" signal, and it provides the simplest expansion. It is
defined as
d@nD = :

1
0

if n = 0
otherwise

and it is plotted in figure 1.2.1, together with a general shifted version d@n - kD.

Unit1.nb

[n k ]

[n ]

Figure 1.2.1: an impulse d@nD and a shifted impulse d@n - kD

Any discrete time signal can be expanded into the superposition of elementary shifted impulses, each
one representing each of the samples. This is expressed as
x@nD =

S x@kD d@n - kD

k=-

where each term x@kD d@n - kD in the summation expresses the n-th sample of the sequence.

x[n]

1. 5

1.0

1.6
1. 2

0.5

0. 5

Figure 1.2.2: Example of a sequence

Example: the sequence shown in figure 1.2.2 can be expanded as
x@nD =
1.5 d@n + 2D - 1.0 d@n + 1D + 1.2 d@nD - 0.5 d@n - 1D + 0.5 d@n - 2D + 1.6 d@n - 3D
This expansion can be viewed as the following animation (double click the figure to animate)

Unit1.nb

1. 5 [ n + 2]

1.5

The significance of this expansion is the fact that any signal, no matter how complicated, is decomposed
into elementary pulses properly scaled in amplitude and shifted in time.
Of particular interest is the Unit Step sequence

u@nD = :

1
if n 0
0 otherwise

shown below. It is a simple exercise to show that

n

u@nD = S d@kD
k=-

u[n]

n
Unit Step Sequence

Homework Problems: Problem 1.1

See if you are following. Click HERE for some Questions (not
graded)

Unit1.nb

1.2.2 Continuous Time Unit Impulse: the Dirac Function

In continuous time we have to be a bit more carefull. A unit impulse d HtL is defined as a function
which is zero for all t 0 , and yet its integral is nonzero. In particular d HtL is such that
d(t)=0 for all t 0,

d HtL dt = d HtL dt = 1.
0+

0-

and it is represented as in figure 1.2.3. It can be viewed as the limit of a sequence of a rectangular signal
1
of width T and height
T , as T 0.

1
T

(t )

T 0

1
T

Figure 1.2.3: Dirac "Delta" Function

Its significance is the fact that for any signal x HtL, continuous at time t, we can write

x Ht - tL d HtL dt = x Ht - tL d HtL dt = x HtL d HtL dt = x HtL

0+

0+

0-

0-

This is known as the "Sifting Property":

x HtL = x Ht - tL d HtL dt = x HtL d Ht - tL dt
+

The rightmost integral in the above expression is easily derived by a change in the integration variable.
Comparing to the equivalent expression in discrete time, which we recall here for convenience,
+

x@nD = S x@kD d@n - kD

k=-

we can see that in both cases we expand a signal (continuous time or discrete time) in terms of a
sequence of unit impulses.
Of particular interest is the Unit Step signal defined as
u HtL = :

shown below. It is a simple exercise to show that

1
if t 0
0 otherwise

Unit1.nb

u HtL = d HtL dt
t

u (t )

t
Unit Step Signal
Homework Problems: Problem 1.2

See if you are following. Click HERE for some Questions (not
graded)

1.2.3 Sinusoidal Signals

As we will be seeing in the Fourier Analysis section, sinusoidal signals are one of the building blocks of
more general signals. For example an audio signal is made of vibrations, of the vocal cords in humans
and animals, or strings or standing waves for musical instruments. When we stand next to a car playing
rock music at a full blast we feel the vibrations right in our own stomach!
A continuous time sinusoidal signal is defined as

x HtL = A cos HW0 t + aL

with A being the Amplitude. If the independent variable t denotes time, and it is measured in seconds,
then Wo indicates the angular frequency in radians per second, and a the phase in radians. Since the
cosinus, as all trigonometric functions, does not change when the argument is shifted by multiples of
2 p , we can see the periodicity of the sinusoidal signal from the expression
x HtL = Acos HW0 t + a + 2 pL = Acos HW0 Ht + T0L + aL = x Ht + T0L

2 p is the period in seconds (if t is in seconds). Amplitude and period are shown in
where T0 =
W0
figure 1.2.4 below.

Unit1.nb

x (t )

A
t

T0
Figure 1.2.4: Sinusoidal Signal showing Amplitude A and Period T0
For example a sinusoid with angular frequency W0 = 2000 p has a period
2p
-3
T0 =
sec = 1 msec.
2000 p = 10
1
The inverse of the period F0 =
T0 is the frequency, and it represents the number of repetitions per
second.

In the example above, the sinusoid has a period T0 = 1 msec, and therefore its frequency is
F0 = 1000 Hz , where the unit Hertz is defined as Hz = sec-1 .
The understanding of frequency and period will be at the basis of a good part of discussions in this
book, and it is summarized here for convenience:
2p
Angular Frequency W0 = 2 pF0 =
T0 radians/sec
1
2p
T0 =
seconds
Period
F0 =
W0
W0
1
Frequency
F0 =
Hz=sec-1
T0 =
2 p

Example: a sinusoid with frequency F0 =250kHz has a period

-3
1
T0 = H
sec = 4 10-6 sec = 4 msec
250 L 10

A discrete time sinusoid is obtained by sampling a continuous time sinusoid with sampling interval Ts ,
as
x@nD = x HnTsL = Acos HW0 Ts n + aL

Unit1.nb

x (t )

x[ n ] = x ( nT s )

Fs =

1
Ts
n

Ts
Figure 1.2.5: A Sampled Sinusoid
By defining the digital frequency w0 = W0 Ts we can write the sampled sinusoid as
x@nD = Acos Hw0 n + aL

rad L sec has

As we can see from the definition, the digital frequency w0 = W0 Ts = 2 p F0 Ts = H
sec
no dimensions. In order to see the meaning of the digital frequency, define the sampling frequency
1
Fs =
Ts , which represents the number of samples taken every second, and write it as
F0
w0 = 2 p
Fs radians

In other words w0 is a relative frequency, with respect to the sampling frequency.

Example: a sinusoid with frequency F0 = 2 kHz is sampled every Ts = 0.1 msec = 10-4 sec. Then
the sampling frequency is Fs = 104 Hz = 10 kHz, and the digital frequency of the sinusoid is
F0
2000
2p
w0 = 2 p
Fs = 2 p
10000 =
5 radians

Homework Problems: Problems 1.3, 1.4

See if you are following. Click HERE for some Questions (not
graded)

Unit1.nb

1.2.4 Complex Exponentials

Although we represent signals in terms of sinusoidal signals (read vibrations), trigonometric functions
do not lend themselves to easy manipulations. Fortunately, a sinusoid can be expanded in terms of complex exponentials, which, on the other hand, exhibit very convenient mathematical properties. In particular recall Euler's formula (or one of the many Euler's formulas):
1
cos HaL =
Hej a + e-j aL
2

1
sin HaL =
Hej a - e-j aL
2j

for any angle a. Therefore, substituting for the appropriate time varying angle, both continuous time and
discrete time sinusoidal signals can be expressed in terms of complex exponential signals as in the
following:
A
A
j a ej2pF0 t +
-j a e-j2pF0 t
x HtL = Acos H2 p F0 t + aL =
2 e
2 e
A
A
x@nD = Acos Hw0 n + aL =
ej a ejw0 n +
e-j a e-jw0 n
2
2

The reason why the exponential signal in general is more attractive than the sinusoidal signal, is because
a number of significant operations we perform on signals become just algebraic manipulations in complex exponentials. For example:
a) Differentiation and Integration. Take any signal x HtL and compute its derivative or its integral
d
with respect to time, as yd HtL =
dt x HtL or yI HtL = x HtL dt . In general both
yd HtL and yI HtL have expressions different from x HtL, UNLESS x HtL is an exponential, in which case differentiation and integration are just algebraic operations as multiplication
and division:
d
d
j2 p F0 t = Hj2pF L ej 2 pF0 t = Hj2pF L x HtL
yd HtL =
0
0
dt x HtL =
dt e
1
j2 p F0 t
1
yI HtL = x HtL dt = ej2 p F0 t dt = I

M
e
=
I

j2pF0
j2pF0 M x HtL

b) Time Shift. Similarly, take any sequence x@nD and shift it in time as y@nD = x@n - LD, with L
being an integer. Again if x@nD is an exponential signal (and ONLY in this case), then the shifted
sequence is obtained just by multiplication, as
y@nD = x@n - LD = ejw0 Hn-LL = He-jw0 LL ejw0 n = He-jw0 LL x@nD

In more technical words we say that the complex exponential signal is an eigenfunction of Differential,
Integral and Time Shift operators. This means that, ONLY in the case of exponentials, these operations
are just algebraic manipulations. All other signals are not so lucky! This is at the basis of most (not all)
the transform techniques (Fourier, Laplace, z-, and all their relatives) which are introduced as tools to
analyze signals, systems and their interactions.

Unit1.nb

Homework Problems:

Problems 1.5, 1.6, 1.7

See if you are following. Click HERE for some Questions (not
graded)

1.2.5 Signal Manipulations

Given a continuous time or a discrete time signal x HtL or x@nD, we define a number of operations,
which we list here for convenience. Apart from obvious algebraic manipulations, like scaling by a
constant, we have to be particularly carefull when we manipulate the time axis. In particular we define:

a) Time Shift: x Ht - t0L or x@n - n0 D , with t0 (real) or n0 (integer) representing a time shift. This
is shown in figure 1.2.6;

x(t t0 )

x(t )

t0

x[n n0 ]

x[n]

n
Figure 1.2.6: Shift in Time

n0

b) Time Scaling: this is easily defined in continuous time as x HatL. However in discrete time, since
the independent variable has to be an index, in general it is not defined, apart from some particular cases
which we are not going to present here. In the continuous time case, according to whether the constant a
is positive or negative, we have two different situations: in the latter (a < 0) the signal is also inverted
on the time axis. This is shown in figure 1.2.7 below.

10

Unit1.nb

x (at )
x (t )

t1

t1
a

t2

t2
a

x (at )

t2
a

t1
a

Figure 1.2.7: Scaling of Time Axis

Homework Problems

a>0
t

a<0
t