Unit-II

Analysis of Discrete System

1
State space description for a linear discrete-
time system:

2
Transformation of discrete-time systems State realization of a polynomial difference system
The polynomial difference system described by

3
Polynomial realization of a state system The state difference system described by

4
State-space to Difference equations

5
So the related difference equation is given by equation

6
 State variable responses of discrete-time linear systems
Linear time-invariant discrete-time models in the standard state equation form

Homogeneous State responses

Note that x(1) = Ax(0) and x(2) =Ax(1), so by substitution we obtain x(2) = A2 x(0). In this way :

7
Properties of discrete-time transition matrix

8
Consider the second order exponential smoother

9
10
 Forced response of discrete-time linear systems
The solution of the in homogeneous discrete-time state equation

11
Assume a0 = 0, a1 = 1/2 , b1 = 1 and b2 = 0 resulting in A as in and B, C, and D as follows

Express the forced response of this system as

12
The system output response of discrete-time linear
systems:

13
 State space to Transfer Function

Consider the third-order system

14
15
 Solving the Discrete-time State Equations
Consider the following state-space and output equations:

16
17
18
19
(ZI-G)-1

20
Note: With the help of GP series you can solve x(k)

y(k)=C x(k)
21
22
 Modeling of sample hold circuit

x(t)
x*(t)
Sampling
x(t)
Make a analog signal to be a discrete signal shown
as in Fig.1 .

x(t) —analog signal .

t
x*(t) —discrete signal . 0 t t t t t t
1 2 3 4 5 6
Ideal sampling switch —sampler Fig1.signal sampling
Sampler —the device which fulfill the sampling.
Another name —the sampling switch — which works like a switch shown as in Fig.2.

T
Some terms
x(t) x*(t)
1. Sampling period T— the time interval of the signal
sampling: T = ti+1 － ti . t t
0 0
Fig2. sampling switch
23
r(t) e(t) e*(t) u*(t) u (t) c(t)
A/D computer D/A process
－

measure
Fig.3 computer control system
Sampling analysis
Expression of the sampling signal:

 
x * ( t )  x( t )   T ( t )  x( t )    (t  kT )   x(kT ) (t  kT )
k 0 k 0
It can be regarded as Fig.4:

24
 Sampling analysis
T
x(t) x*(t)

x(t) x*(t)
δT(t)
× ＝
t t t
0 0 0
Modulation modulating modulated
signal pulse(carrier) wave
Fig.4 sampling process
  
1
because :  T ( t )    (t  kT )   C ne jk s t

T
 e jk st   s  2 / T
k 0 k   k  
 T /2
1 
0
1  1 
0
C  1 jk s t jk s t
 n T   T (t )e dt  T   T (t )e dt  T   T (t )dt  T 
 T / 2 0 0 
We have:

25
 Sampling analysis
 Laplace 
1 jk s t 1
x * ( t )  x( t ) 
T
e  X * ( s ) 
T
 X ( s  jk s )
k   transforma tion k  

  
 here : X ( s )  L[ x( t )]  x ( t )e  st dt 
  
 0 

1
The frequency spectrum of x * ( t ) : X* ( j ) 
T
 X [ j(  k s )]
k  
This means: for the frequency spectrum of x(t) shown in Fig.5, the frequency spectrum
of x*(t) is like as Fig..6.

X * ( j )
X ( j )  max
Filter Only:  s  2max
  X ( j ) could be
 max  max reproduced
s
 max
Fig.5 2 s
Fig.6 26
 Sampling theorem ( Shannon’s theorem)

If the analog signal could be whole restituted from the sampling signal, the sampling
s
frequency must be satisfied :


 s  2 max or T 
 max
here :  max  the maximu m frequenc y of the analog signal .
T  sampling period .
 s  sampling frequency ,  s  2 T .
Zero-order hold

Usually the controlled process require the analog signals, so we need a discrete-to-
analog converter shown in Fig.7.

x*(t) discrete-to-analog xh(t) Fig.7

converter D/A convert
27
 Zero-order hold
The ideal frequency response of the D/A converter is shown in Fig.8.

To put the ideal frequency response in A(ω)  ( )

practice is difficult, the zero -order hold ω
is usually adopted. Fig.8
The action of the zero-order hold is shown
in Fig.9. x*(t)
x(t)
xh(t)
The mathematic expression of xh(t) :

xh (t )  x(kT ) kT  t  (k  1)T
The unity pulse response of the zero-order hold is Fig.9
shown in Fig.10.
g(t)
The transfer function of the zero-order hold can be
obtained from the unity pulse response:
Ts t
1 e
G( s )  L1( t )  1( t  T )  T
s Fig.10
28
Mathematical modeling of the sampling systems
Z-transfer (pulse) function

Definition: Z-transfer (pulse) function — the ratio of the Z-transformation of the

output signal versus input signal for the linear sampling systems in the zero-initial
conditions, that is:

C(z)
G( z ) 
R( z )
1. The Z-transfer function of the open-loop system

r(t) c*(t)
G1(s) G2(s) c(t)
T T
R(z) C(z)
G1(z)G2(z) G1(z) =Z [ G1(s)] G2(z) =Z [ G2(s)]
c*(t)
r(t)
G1(s) G2(s) c(t)
T

R(z) C(z)
G1G2(z) G1G2(z) =Z [ G1(s)G2(s) ]
29
 Z-transfer (pulse) function

2. The z-transfer function of the closed-loop system

r c
 GH ( z )  Z G( s ) H ( s )
R( z )G( z )
－
G(s)
C(z) 
H(s)
1  GH ( z )

 RG( z )  Z R( s )G( s )

r c RG( z )
G(s) C(z) 
－ 1  GH ( z )
H(s)
r c R( z )G ( z )
G(s) C(z) 
－ 1  G( z ) H ( z )
H(s)
r c RG1( z )G2 ( z )
－
G1(s) G2(s) C(z) 
1 G1G2 H ( z )
H(s)
r c R( z )G1( z )G2 ( z )
－
G1(s) G2(s) C(z) 
1 G1( z )G2 H ( z )
H(s)

30
Z-transfer (pulse) function

r c
G1(s) G2(s) G3(s)
－
H(s)
RG1 ( z )G2 ( z )G3 ( z )
C(z) 
r c
1  G2 ( z )G3 ( z )G1 ( z ) H ( z )
G1(s) G2(s)
－ －
H1(s)
H2(s)
R( z )G1 ( z )G2 ( z )
C (z) 
1  G2 H1 ( z )  G1 ( z )G2 H 2 ( z )
r c
G1(s) G2(s) G3(s)
－ －
H1(s)
H2(s)

RG1 ( z )G2 ( z )G3 ( z )

C(z) 
1  G2 ( z )G3 H1 ( z )  G2 ( z )G1G3 H 2 ( z )
31
 Time-domain analysis of the sampling systems

The stability analysis

1. The stability condition
The characteristic equation of the sampling control systems:

1  GH ( z )  0
∵ z  eTs  1  GH (eTs )  0
Suppose:
s    j  eTs  eT (  j )  eT  e jT
In s-plane, α need to be negative for a stable system, it means:

eTs  z  eT  1
So we have:
The sufficient and necessary condition of the stability for the sampling control
systems is:
The roots zi of the characteristic equation 1+GH(z)=0 must all be inside the unity circle
of the z-plane, that is:
zi  1
32
The stability analysis critical stability Im
z-plane
The graphic expression of the stability
1 Re
condition for the sampling control systems
is shown in Fig..

unstable zone Stable zone

2. The stability criterion
Fig.
In the characteristic equation 1+GH(z)=0, substitute z with
w 1 —— W (bilinear) transformation.
z
w 1

 Proof : suppose w    j , z  x  jy , then : 

 
 z  1 x  jy  1  x  1  jy  x2  y2  1 2y 
 w    j   
 
  j 
 z  1 x  jy  1  x  1  jy  ( x  1) 2
 y 2
( x  1) 2
 y 2

 
   0  x 2
 y 2
 1  0  x 2
 y 2
1 
 ( for the left half of the w-plane )  ( inside the unit circle of the z-plane ) 
 33

 Time-domain analysis of the sampling systems
The unit-step response analysis
A( z ) z
suppose : C ( z )   ( z ) R( z )  
( s  p1 )( s  p2 )    ( s  pn ) z  1
A0 z n Ai z
 
z  1 i 1 z  pi
n kT Im
then : c( kT )  A0   Ai ( pi )T
i 1
1
Analyzing c(kT) we have the Re
graphic expression of C(kT) is shown
in Fig..

Fig.
34
 The Routh-Hurwitz criterion
• Give the characteristic equation Q(z)
Z-plane  -plane

1  (T 2)
Z
1  (T 2)
2 Z 1
or  
T Z 1 Q( )  Q( z ) z  (1 T  ) /(1 T  )
2 2

The system is stable all The LTI discrete-time system is stable

the roots of Q(z)=0 must lie all the roots of Q(w)=0
with in the unit circle have negative real parts.
The Routh-Hurwitz stability
criterion can be used

35
 Jury’s Stability Test

This stability criterion for LTI discrete-time systems is similar to

the Routh-Hurwitz criterion for LTI continuous-time systems
and can be directly applied to examine the characteristic equation.
Described by Q( z )  an z n  an -1 z n -1      a1 z  a0  0 an  0

36
the array is formed as
z0 z1 z 2  z n - k  z n -1 zn
a0 a1 a2  an - k  an -1 an
an an -1 an -2  ak  a1 a0
a0 an - k
b0 b1 b2  bn - k  bn -1  bk 
an ak
bn -1 bn -2 bn -3  bk -1  b0
b0 bn - k
c0 c1 c2  cn - k   ck 
bn -1 bk
cn -2 cn -3 cn -4  cn -2
   
l0 l1 l2 l3
l3 l2 l1 l0
m0 m1 m2

37
The necessary and sufficient conditions for Q(z)=0
to have all the roots inside the unit cycle are.
Q(1)  0, (1) n Q(1)  0, a0  an ,
b0  bn -1 , c0  cn -2 , d 0  d n -3  m0  m2 .
For an nth-order system, Jury's test may be applied in
the following manner:
1. Check the three conditions
Q(1)  0,( 1) n Q( 1)  0, and a0  an
which requires no calculations. Stop if any of
these conditions are not satisfied.
2. Construct the array, checking the stability conditions as each row
is calculated. Stop if any one condition is not satisfied
38
Example
Consider a discrete-time system whose characteristic equation is given as
0.368 z  0.264
1  KG ( z )  1  ( ) K , Find k such that the system is stable.
z -1.368 z  0.368
2

Answer:
Q( z )  z 2  (0.368K  1.368) z  (0.368  0.264 K )  0

The Jury array is z0 z1 z2

0.368  0.264 K 0.368K  1.368 1
Stability conditions 

Q(1)  0.632 K  0 K 0 

2.736 
(1) 2 Q(1)  2.736  0.104 K  0  K   26.3  Thus 0  K  2.39
0.104 
0.632 
0.368  0.264 K  1 K  2.39 
0.264 

39
Example: Suppose that the 3rd-order characteristic equation

Q( z )  z 3  1.8 z 2  1.05 z  0.20  0

(1)The first stabzility conditions require that
Q(1)  z 3  1.8 z 2  1.05 z  0.20  0.05  0
z 1

( 1)3 Q( 1)  [1  1.8  1.05  0.2]  4.05  0

a0  0.20  0.20  1
(2)The Jury array is calculated to be
z0 z1 z2 z3
 0.20 1.05  1.8 1  0.20  1
1  1.8 1.05  0.20
 0.96 1.59  0.69  stop  0.96  0.69

40
Therefore,since all the conditions are satisfied,
the system is asymptotically or BIBO stable,
In fact
Q( z )  ( z  0.5)2 ( z  0.8)
has all its roots inside the unit circle.
Note that
0.20 1 0.20 1.8 0.20 1.05
0.96  ,1.59  ,0.69 
1 0.20 1 1.05 1 1.8

41