Bibo

1.
External (BIBO) Stability of LTI Systems
If every Bounded Input produces Bounded Output,

system is externally stable
equivalently, system is BIBO stable
|g(n)| < BIBO Stability
n=
Dont care about what unbounded input does...
CL 692 Digital Control, IIT Bombay
2.
c
Kannan
M. Moudgalya, Autumn 2006
Condition for BIBO Stability
Is it possible for system to be

P
BIBO stable and
|g(i)| ?
|g(n)| < BIBO Stability
n=
y(n) =
Produce the following signal:

( g(k)
g(k) 6= 0,
u(k) = |g(k)|
0
g(k) = 0.
g(k)u(n k)
k=
Suppose
n= |g(n)|
< .
Bounded u with |u(n)| < M
{u(n)} = {. . . , 1, 1, 1, 1, 1, . . .}
Evaluate one term:

|y(n)|
y(0) =
|g(k)||u(n k)|
k=
|g(k)|
k=
X
k=
X
k=
X
k=
As M is bounded, so is y(n). This is true for

all n.
u(k)g(n k)|n=0
u(k)g(k)
X
[g(k)]2
=
|g(k)|
|g(k)|
k=
So, not BIBO stable
c
Kannan
3.
Z-transform - Motivation
X
y(n) =
u(k)g(n k)
Also carryout multiplication:
k=
{u} = {u(0), u(1), u(2)}

{g} = {g(0), g(1), g(2)}
(u(0) + u(1)z 1 + u(2)z 2)

(g(0) + g(1)z 1 + g(2)z 2) =
Let these be zero at all other times.

y(0) = u(0)g(0)
y(1) = u(0)g(1) + u(1)g(0)
y(2) = u(0)g(2) + u(1)g(1) + u(2)g(0)
y(3) = u(1)g(2) + u(2)g(1)
u(0)g(0)+
(u(0)g(1) + u(1)g(0))z 1+
(u(0)g(2) + u(1)g(1) + u(2)g(0))z 2+
(u(1)g(2) + u(2)g(1))z 3+
Think of z as a position marker - coefficient of z i occurs at ith instant

u(0) + u(1)z 1 + u(2)z 2 - a way of representing a sequence with three
terms: {u(0), u(1), u(2)}
The z polynomial can often be represented by a compact expression even
if the original sequence has infinitely many nonzero terms
3
4.
c
Kannan
Two Sequences to Show Importance of Convergence

u1 (n) = an 1(n)
u2 (n) = an 1(n 1)
0
1
0.9
0.5
0.8
0.7
0.9 1(n1)
0.5
0.9 1(n)
0.6
0.4
1.5
0.3
0.2
2.5
0.1
0
10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
a =
n =
y =
for
5
Time (n)
10
15
0.9;
10:20;
zeros ( s i z e ( n ) ) ;
i = 1 : length ( n )
i f n ( i )>=0,
y ( i ) = a n ( i ) ;
end
end
A = axes ( F o n t S i z e , 1 4 ) ;
s e t ( get (A , X l a b e l ) , F o n t S i z e , 1 4 ) ;
o = stem ( n , y ) ;
s e t ( o ( 1 ) , Marker , . ) ;
x l a b e l ( Time ( n ) ) ;
y l a b e l ( 0 . 9 n1 ( n ) ) ;
3
10
20
1
2
3
4
5
6
7
a =
n =
y =
for
5
Time (n)
10
15
20
0.9;
10:20;
zeros ( s i z e ( n ) ) ;
i = 1 : length ( n )
i f n ( i )<=1,
y ( i ) = (a n ( i ) ) ;
end
14
end
A = axes ( F o n t S i z e , 1 4 ) ;
s e t ( get (A , X l a b e l ) , F o n t S i z e , 1 4 ) ;
o = stem ( n , y ) ;
s e t ( o ( 1 ) , Marker , . ) ;
x l a b e l ( Time ( n ) ) ;
y l a b e l ( 0.9 n1(n1) ) ;
c
Kannan
8
9
10
11
12
13
5.
Z-transform - Convergence Condition for Uniqueness

u1(n) = an1(n)
u2(n) = an1(n 1)
X
U2(z) =
an1(n 1)z n
Take Z-transform:
X
U1(z) =
anz n
=
n=0
n=
a z
n=1
(az 1)n
=1
n=0
n n
amz m
m=1
(a1z)m
m=0
If the sum exists,

z
1
=
=
1 az 1 z a
If the sum exists,
z
1
=
1 a1z z a
Any difference? Clue: If the sum exists. Convergence Condition:
=1
az 1 < 1
a<z
a1z < 1
z<a
6.
c
Kannan
Z-transform - Convergence Condition for Uniqueness

u1(n) = an1(n)
z
U1(z) =
za
1
az < 1
u2(n) = an1(n 1)
z
U2(z) =
za
1
a z<1
May want a complex, would want z complex

So make the convergence absolute
|az 1| < 1
|a| < |z|
|a1z| < 1
|z| < |a|

c
Kannan
7.
Z-transform Definition
Z-transform of a sequence {u(n)}, denoted by U (z), calculated using:

U (z) =
u(n)z n
n=
where z is such that

U (z) =
|u(n)z n| <
n=
Stronger condition of absolute convergence allows u and z to be complex.

The main advantage in taking Z-transform is that cumbersome convolution calculations can be done easier.
Suitable for studying stability, causality and oscillatory behaviour of even
infinitely long signals. Will use this fact while designing controllers.
There is an added burden of converting the polynomial to a closed form
expression and inverting this procedure to arrive at the resultant polynomial. It is possible to develop techniques to simplify these operations.
8.
c
Kannan
Region of Convergence (ROC) - Property 1
ROC consists of a ring centred around the origin

z0
|z0 |
|z|
z
u(n)z n
X
z0 ROC
|u(n)z0n| <
X
X
X
|z|=|z0 | X
>
|u(n)z0n| =
|u(n)||z0n| =
|u(n)||z0|n =
|u(n)||z|n
X
=
|u(n)z n|
U (z) =
Thus, arbitrary z also belongs to ROC
c
Kannan
9.
ROC Property 2 - Poles are Outside ROC

ROC does not have any pole
At a pole, transfer function is infinite, does not converge
10.
c
Kannan
ROC - 3: Larger Circles are in ROC for Causal Signals
For sequence {u(n)}, with u(n) = 0 for n < 0, if z0 ROC, all z such that
|z| > |z0| will belong to ROC
>
z0
|z0 |
|z|
z
=
U (z) =
u(n)z n
>
n=
z0 ROC
|u(n)z0n| <
n=0
X
n=0
X
n=0
X
n=0
X
n=0
|u(n)z0n|
|u(n)||z0n|
n
|u(n)||z0|
X
|u(n)|
n=0
|u(n)|
,
|z|n
|z0|n
|z| > |z0|
|u(n)z n|
n=0
Thus, arbitrary z also belongs to ROC

10
c
Kannan
11.
ROC - 4: Poles of Causal, Stable Systems in Unit Circle

Thus, there is absolute convergence at |z| = 1.
Causal and stable system

Impulse response g(n)
Equivalently, the unit circle belongs to the region of convergence.
Z-transform of g(n), namely

G(z), will have poles inside
unit circle
As it is a causal sequence, by
Property 3, all points outside
the unit circle also belong to
the ROC
As g(k) is causal and stable,
|g(k)| <
k=0
G(z) =
>
X
k=0
X

g(k)z k
|g(k)| =
k=0
z=1
k=0
12.
From property 2, the poles

cannot lie in ROC - they have
to lie within the unit circle
g(k)z k
11
c
Kannan
ROC - 5: Z-transform of Causal Signal is Proper
u(k) is a causal sequence

U (z) =
N (z)
D(z)
with
N (z) is a polynomial of degree n

D(Z) is a polynomial of degree m
n m.
Since u(k) is causal, by property 3, z = is in ROC

If n > m, U (z) will diverge at
Thus n m.
12
c
Kannan
13.
Z-transform Theorems - Linearity

Z [{u1(n)} + {u2(n)}] = U1(z) + U2(z)
where,
u1(n) U1(z)
u2(n) U2(z).
Here and are arbitrary scalars.
LHS =
=
X
n=
[u1(n) + u2(n)] z n
u1(n)z
n=
u2(n)z n
n=
= U1(z) + U2(z) = RHS
14.
13
c
Kannan
Example 1 - Linearity
Find the Z-transform of

u1(n) = 2(n) 3(n 2)
+ 4(n 5)
X
U1(z) = 2
(n)z n
3
+4
n=
X
n=
Find the Z-transform of

{u2(n)} = [2 + 4(3)n] {1(n)}
X
U2(z) =
[2 + 4(3)n] z n
n=0
4z
2z
+
, |z| > 1 |z| > 3
z1 z+3
6z 2 + 2z
, |z| > 3
=
(z 1)(z + 3)
=
(n 2)z n
(n 5)z n
n=
2
= 2 3z + 4z 5
z 1 finite
2z 5 3z 3 + 4
=
|z| > 0.
z5
14
c
Kannan
15.
Example 2 - Linearity
Find the Z-transform of cos wn 1(n)

and sin wn 1(n).
cos n + j sin n = ejn

Z [cos n + j sin n] = Z ejn
X
jn

Z e 1(n) =
ejn z n 1(n)
n=
X
jn
z n
n=0
Comparing real and imaginary parts,

Z [cos n] = Re Z ejn

Z [sin n] = Im Z ejn .
z(z cos )
(cos n)1(n)
(z ej )(z ej )
z sin
(sin n)1(n)
(z ej )(z ej )
z
, |z| > |ej | = 1
j
ze
z(z ej )
=
(z ej )(z ej )
z(z cos + j sin )
=
(z ej )(z ej )

16.
z(z cos )
z sin
+j
j
j
j
(z e )(z e )
(z e )(z ej )

15

c
Kannan
Z-transform - Shifting
Example:
Z [u(k + d)] = z d U (z)
X
Z [u(k + d)] =
u(k + d)z k
k=
X
d
=z
If
{u(n)} U (z),
u(k + d)z (k+d)
then
k=
{u(n + 3)} z 3 U (z)

{u(n 2)} z 2 U (z)
= z U (z)
16
c
Kannan

Bibo

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Bibo

Încărcat de

Drepturi de autor:

Formate disponibile

1.

External (BIBO) Stability of LTI Systems

If every Bounded Input produces Bounded Output,

|g(n)| < BIBO Stability

Dont care about what unbounded input does...

CL 692 Digital Control, IIT Bombay

Condition for BIBO Stability

Is it possible for system to be

|g(n)| < BIBO Stability

Produce the following signal:

Bounded u with |u(n)| < M

Evaluate one term:

As M is bounded, so is y(n). This is true for

CL 692 Digital Control, IIT Bombay

So, not BIBO stable

{u} = {u(0), u(1), u(2)}

(u(0) + u(1)z 1 + u(2)z 2)

Let these be zero at all other times.

Think of z as a position marker - coefficient of z i occurs at ith instant

CL 692 Digital Control, IIT Bombay

Two Sequences to Show Importance of Convergence

Z-transform - Convergence Condition for Uniqueness

If the sum exists,

If the sum exists,

CL 692 Digital Control, IIT Bombay

Z-transform - Convergence Condition for Uniqueness

May want a complex, would want z complex

CL 692 Digital Control, IIT Bombay

Z-transform of a sequence {u(n)}, denoted by U (z), calculated using:

where z is such that

Stronger condition of absolute convergence allows u and z to be complex.

Region of Convergence (ROC) - Property 1

ROC consists of a ring centred around the origin

Thus, arbitrary z also belongs to ROC

CL 692 Digital Control, IIT Bombay

ROC Property 2 - Poles are Outside ROC

At a pole, transfer function is infinite, does not converge

CL 692 Digital Control, IIT Bombay

ROC - 3: Larger Circles are in ROC for Causal Signals

|z| > |z0|

Thus, arbitrary z also belongs to ROC

ROC - 4: Poles of Causal, Stable Systems in Unit Circle

Causal and stable system

Equivalently, the unit circle belongs to the region of convergence.

Z-transform of g(n), namely

As g(k) is causal and stable,

CL 692 Digital Control, IIT Bombay

From property 2, the poles

ROC - 5: Z-transform of Causal Signal is Proper

u(k) is a causal sequence

N (z) is a polynomial of degree n

Since u(k) is causal, by property 3, z = is in ROC

CL 692 Digital Control, IIT Bombay

Z-transform Theorems - Linearity

= U1(z) + U2(z) = RHS

CL 692 Digital Control, IIT Bombay

Find the Z-transform of

Find the Z-transform of

Find the Z-transform of cos wn 1(n)

Comparing real and imaginary parts,

CL 692 Digital Control, IIT Bombay

Z [u(k + d)] = z d U (z)

u(k + d)z (k+d)

{u(n + 3)} z 3 U (z)