UEEA 2183

DIGITAL SIGNAL PROCESSING

JAN 2016

TOPIC 2
Discrete-Time Signal and System

Frequency of Sinusoidal Sequences

Consider 1 = cos 1 + and 2 = cos 2 +
with
0 1 < 2
and
2 2 2 + 1
where k is any positive integer.
If
2 = 1 + 2,
then 2 = cos 2 +
= cos 1 + 2 +
= cos 1 + + 2
= cos 1 +
= 1

Frequency of Sinusoidal Sequences

Thus, 1 = cos is the same as
2 = cos + 2 where k is an integer.
It is similar to cos = cos + 2 , i.e., a cosine function
is the same when it is shifted by integer values of 2.
E.g., the signal 1 = cos 0.1 is the same as the
signal 2 = cos 2.1 = cos 0.1 + 2 .
E.g. 3 = cos 1.9 = cos 2 0.1
= cos 2 cos 0.1 + sin 2 sin 0.1
= cos 0.1 = 1
4

Frequency of Sinusoidal Sequences

Consider 1 = cos 1 and 2 = cos 2
with
0 1 <
and 2 2.
Let 2 = 2 1 ,
then 2 = cos 2 1 = cos 1 = 1
Hence, a sinusoidal sequence with 2 in the range
2 2 assumes the identity of a sinusoidal
sequence with 1 = 2 2 in the range 0 1 < .
The frequency is called the folding frequency.
5

Frequency of Sinusoidal Sequences

Frequency of Sinusoidal Sequences

The frequency of oscillation of = A cos
increases as increases from 0 to , and then the
frequency of oscillation decreases as increases from
to 2.
Therefore, for discrete-time sinusoidal sequence, the
highest frequency possible is = radians/sample. At
this frequency, there are 2 samples for 1 complete cycle.
7

Frequency of Sinusoidal Sequences

The time variable, t of the continuous-time signal is
related to the time variable, n of the discrete-time signal
only at discrete-time instant, tn given by
2
= = =

where Ts = sampling interval

1
fs = sampling frequency, =

= sampling angular frequency, = 2

Frequency of Sinusoidal Sequences

Given = sin 2 = sin .
The corresponding discrete-time signal is
2
= sin = sin
= sin

where =

= is the normalized angular frequency

of (unit is radians per sample).

Analog sinusoidal functions with different frequencies may
have the identical sampled signals with same period.

Frequency of Sinusoidal Sequences

E.g. sin 0.3 and sin 0.1 have the same period
2
20
2
20
20Ts
=

=
, even though sin 0.3
0.3

0.1

has a rate of change 3 times as fast as sin 0.1 .

E.g. Consider 3 sequences generated by uniformly
sampling the 3 cosine functions of frequencies 3 Hz, 7 Hz
and 13 Hz, respectively with sampling rate of 10 Hz.
1 = cos 6 ,
1 = cos 0.6
2 = cos 14 ,
2 = cos 1.4
3 = cos 26 ,
3 = cos 2.6

10

Frequency of Sinusoidal Sequences

1 = solid line, 2 = dashed line, 3 = dashed-dot line

11

Frequency of Sinusoidal Sequences

From the plots, each sequence has exactly the same
sample value for any given n.
It can be verified that
2 = cos 1.4 = cos 2 0.6 = cos 0.6
3 = cos 2.6 = cos 2 + 0.6 = cos 0.6
Therefore, 1 , 2 and 3 are identical.
A continuous-time sinusoidal signal of higher frequency
acquiring the identity of a sinusoidal sequence of lower
frequency after sampling is called aliasing.
12

Frequency of Sinusoidal Sequences

Given = cos 2 and = cos 2 .
In general, the family of continuous-time sinusoids
, = cos 2 + 2 ,
k = 0, 1, 2,
leads to identical sampled signals:
2
, = cos 2 +

= cos 2 + 2
= cos 2
=

13

Frequency of Sinusoidal Sequences

E.g. sin 0.1 with sampling period Ts = 1 (fs = 1) can
have the following continuous-time sinusoids:
sin + 2 = sin 0.1 + 2 1
= sin 0.1 2
when k = -1
= sin 0.1
when k = 0
= sin 0.1 + 2
when k =1
which are sin 0.1 , sin 2.1 , sin 1.9 and so on.

0.1
where = =
= 0.1

Since 0.1 has the smallest magnitude, the frequency of

14
the digital sinusoid sin 0.1 is 0.1 rad/s.

Frequency of Sinusoidal Sequences

Since

where
= digital frequency
= analog frequency

Relationship of the frequencies between digital and

Digital
analog sinusoid.
/T
__

Ts

__
- 2
T
s

0
- __
Ts

- __
Ts

__

Ts

Analog
2
__
Ts

15

Frequency of Sinusoidal Sequences

Example 1
Find the frequency of the digital signal sin 4.4
sampling period is Ts = 0.1.
Solution
sin 4.4 = sin 4 + 0.4 = sin 0.4
=
0.4
= =
= 4 /

0.1

if the

16

Frequency of Sinusoidal Sequences

Example 2
Determine the discrete-time signal v[n] obtained by
uniformly sampling a continuous-time signal va(t) composed
of a weighted sum of 5 sinusoidal signals of frequencies
30 Hz, 150 Hz, 170 Hz, 250 Hz and 330 Hz, at a sampling
rate of 200 Hz, as given below:
= 6 cos 60 + 3 sin 300 + 2 cos 340
+ 4 cos 500 + 10 sin 660
17

Frequency of Sinusoidal Sequences

Solution
= =

200

= 6 cos 0.3 + 3 sin 1.5 + 2 cos 1.7

+ 4 cos 2.5 + 10 sin 3.3
= 6 cos 0.3 + 3 sin 2 0.5 + 2 cos 2 0.3
+ 4 cos 2 + 0.5 + 10 sin 4 0.7
= 6 cos 0.3 3 sin 0.5 + 2 cos 0.3
+ 4 cos 0.5 10 sin 0.7
= 8 cos 0.3 + 5 cos 0.5 + 0.6435 10 sin 0.7

18

Frequency of Sinusoidal Sequences

The components
3 sin 1.5 , 2 cos 1.7 , 4 cos 2.5 , 10 sin 3.3 have
been aliased into the components
3 sin 0.5 , 2 cos 0.3 , 4 cos 0.5 , 10 sin 0.7 .
The resulting discrete-time sequence composed of only 3
sinusoidal sequences of : 0.3, 0.5 and 0.7.
19

Frequency of Sinusoidal Sequences

Identical discrete-time signal is also generated by
uniformly sampling at a 200 Hz sampling rate the
following continuous-time signals:
= 8 cos 60 + 5 cos 100 + 0.6435
10 sin 140
and
= 2 cos 60 + 4 cos 100 + 10 sin 260
+ 6 cos 460 + 3 sin 700

20

Linear Systems
A linear system obeys the homogeneity and superposition
principles.
Homogeneity means if the system output is given
input , then given input , the output is a .
Superposition principle states that if input 1 and
2 produces output 1 and 2 respectively, then
an input of 1 + 2 produce an output 1 + 2 .
Thus a linear system fed with the input a1 + 2 will
produce the output a1 + b2 .
21

Linear Systems
A useful property of linear systems is frequency
preservation. If the input contains several frequencies,
then the output will contain only those frequencies.
Non-linear system does not preserve frequencies.
E.g. = 2 . With = sin ,
1 cos 2
2
= =
2
which does not contain the original frequency .
22

Time Invariant Systems

A time invariant system is one where system properties do
not change with time.
The only effect of time-shifting an input will just result in a
corresponding time-shift of the output.
An example of a time-invariant system with output
= + 2 1
If the input is delayed by 0 , then the output is
0 + 2 0 1 = 0
which is the output delayed by 0 .

23

Time Invariant Systems

An example of a time-dependent system with output
= 1 + 0.5
If the input is delayed by 0 , then the output is
0 1 + 0.5 0 0
Time invariant system also preserves frequencies.
Time-shifting the input frequency components will only
affect the phase of the output components.

24

Time Invariant Systems

Linear Time-Invariant (LTI) systems are easy to analyze,
since Fourier techniques allow us to decompose any
signals into sums of sinusoids. The output is a weighted
sum of the same sinusoids shifted in phase.
The time-invariance property ensures that for a specified
input, the output of the system is independent of the time
the input is being applied.

25

Other System Properties

LTI systems also possess the properties of association
and commutation.
The associative property allows us to break a big
system into smaller ones and analyse separately.
The commutative property means that the order of
cascaded subsystems in a large system can be
rearranged without affecting the final output.
Causal systems are those where there is no output until
the application of an input.
26

Other System Properties

Thus the present output of causal systems only depends
on present and past samples of the input.
Changes in output samples do not precede changes in
the input samples.
A stable system is one which produces a finite or
bounded output if the input is bounded.
Such systems are sometimes described as bounded input
bounded output (BIBO).
27

Other System Properties

E.g. Given an output

= +1 1 + ,
=0

where 1 = initial condition.

If > 1,
lim =

The system is unstable since the output is unbounded

given a bounded input.
28

Other System Properties

E.g. Given the output of the M-point moving average filter
1

1
=

=0

For a bounded input <

1

1

=

=0

1
1
<
<

=0

indicating that the system is BIBO stable.

29

Other System Properties

An invertible system is one where can be used to
determine .
An example of a non-invertible system is one with
= 2 . If output is 4, input can be 2.
A memoryless system is one where the output depends
only on the present input.

30

Digital Signals as Sum of Impulses

Any digital signal can be represented as a series of
impulses,
= + 2 + 2 + 1 + 1 + 0
or

+ 1 1 + 2 2 +

For causal signals,

=
=

=
=0

31

Digital Signals as Sum of Impulses

In graphical form, the signal

-1

x[n]

-1

x[-1] [n+1]

1
x[0] [n]

-1

is just a sum of the various

weighted impulses:

x[1] [n-1]

-1

n
32

Impulse Response
When the unit impulse, is inputted into an LTI system,
the output is called the natural or impulse response,
of the system.

LTI
System
0

33

Impulse Response
Example 3
Work out the first 4 sample values of the impulse response
from the filter shown.
x[n]
y[n]
+
_
-0.9

z-1

Solution
Here,
= 0.9 1 +
By setting = , the impulse response is
= 0.9 1 +

34

Impulse Response
The system is causal. Therefore,
0 = 0.9 1 + 0 = 0 + 1 = 1

1 = 0.9 0 + 1 = 0.9 1 + 0 = 0.9

2 = 0.9 1 + 2 = 0.9 0.9 + 0 = 0.81

3 = 0.9 2 + 3 = 0.9 0.81 + 0 = 0.729

The impulse response oscillates about zero and decays as
time passes by.
35

Digital Convolution
Consider the digital signal shown:

2
0

x[n]

n
-1

1
0

h[n]
n

-1

Suppose it is passed through an LTI system with impulse

response shown:

36

Digital Convolution
The output due to the first non-zero component of is
weighted by 1 and shifted left one step because
of homogeneity and time invariance properties of LTI
systems.
The next output due to the second non-zero component of
is weighted by 0 with no time shift.
1

-1

x[-1] h[n+1]

x[0] h[n]

0
-2

37

Digital Convolution
Similarly the output due to 1 is 1 1 , and the
output due to the last non-zero input component is
2 2 .
Since the LTI system also obeys the superposition
principle, the total non-zero output due to is
= 1 + 1 + 0 + 1 1 + 2 2

38

Digital Convolution
1

x[-1] h[n+1]

-1
4

-1

x[0] h[n]
n

x[2] h[n-2]

1
-2
7

-2
6

3
1 1

x[1] h[n-1]
n

y[n]
0

0
-2

-3
39

Digital Convolution
In general, the output due to an input from an LTI system
with impulse response is just

=
=

This is called the digital convolution of with . It is

done by flipping the impulse response (time-reversing it), then
shifting, multiplying it with and summing the products.
Thus the output of any digital LTI system is just the digital
convolution of the input with the systems impulse response.
40

Digital Convolution
Example 4
The impulse response of an LTI system is
0 = 2; 1 = 1; 2 = 1
and = 0 for all other .
Input is:
1 = 1; 0 = 2; 1 = 3; 2 = 1
and zero elsewhere.
Find the output 2 .

41

Digital Convolution
Solution

2 = 2
=1

= 1 3 + 0 2 + 1 1 + 2 0

= 1 0 + 2 1 + 3 1 + 1 2

= 1

42

Averaging Systems
Suppose we have a signal which fluctuates greatly, like
stock prices, and we want to smooth out the signal.
We can do this by using a moving average filter. For
example, a 5-point moving average filter will output
= 0.2 + 2 + + 1 + + 1 + 2

Any short-term fluctuations will be made smaller because

only 20% of each sample value contributes to the output.
43

Averaging Systems
This filter can be obtained by time convolving the input
with the impulse response
0.2 2 2
=
0

In general, if we want a 2 + 1 point moving average

filter, we can design the filter to have the impulse response
1/ 2 + 1
=
0

44

Averaging Systems
Note that the filter is non-causal since it requires future
inputs for present output.
This is not a problem if we pre-record the data first, or we
delay the output until all the inputs required are available.

45

Properties of Convolution
Commutative: =

Given the definition of the convolution sum:

= =
=

Substituting = , we get

= =
=

46

Properties of Convolution
Distributive: 1 + 2
1 + 2

= 1 + 2

= 1 + 2
=

= 1 + 2

= 1 + 2

47

Properties of Convolution
Associative:

=
=

= =

48

Properties of Convolution
Substituting = ,

= =

=
=

49

Properties of Convolution
Two subsystems in parallel is equivalent to a single
system with impulse response the sum of the 2
subsystem impulse responses.

50

Transient Response
When a signal is applied to a system, the output will
experience a start-up transient before settling to the
steady-state response.
A stop transient is also generated when the input is
removed.
Transients are important because they mask the steadystate response, and may also be large enough to damage
the system.
51

Transient Response
If a system is described with a difference equation

=0

=0

where = input of the system

= output of the system
, = constants
max(N, M) = order of the discrete-time system
(order of the difference equation)
52

Transient Response
If the system is assumed to be causal, then

=1

=0

= +
0
0

provided 0 0.
The output response consists of 2 components
= +
where = complementary or homogeneous solution
= particular solution

53

Transient Response
The transient or natural response is given by the
homogeneous solution or zero-input ( = 0) response.

= 0

=0

The steady-state or forced response is given by the

particular solution ( 0 ) or zero-state response
(response with initial state of zero).
54

Transient Response
The transient response is caused by the non-zero initial
conditions in the system decaying to zero.
The steady-state response is due entirely to the input
signal.
Convolving the input with the impulse response of an LTI
system produces the total response from the system.

55

Transient Response
Example 5
A causal FIR discrete-time system is characterized by an
impulse response = 4, 5,6, 3 , 0 3 . Its
output samples for an input are then computed using
= 4 5 1 + 6 2 3 3
For a unit step sequence input,
0 = 4 0 = 4
1 = 4 1 5 0 = 1
2 = 4 2 5 1 + 6 0 = 5

56

Transient Response
3 = 4 3 5 2 + 6 1 3 0 = 2
4 = 4 4 5 3 + 6 2 3 1 = 2
5 = 4 5 5 4 + 6 3 3 2 = 2

It follows from the above that = 2 for 3, or the

output has reached the steady-state at = 3.
The output samples for = 0,1,2 are composed of the
samples of the transient and the steady-state responses.
57