Chapter 8
Transfer Function (TF) to State Space (SS) Form
Phase Variable Form
T(s) =
=
Masons Rule pg. 84-85
5s2 + 4s 12
s 3 + 6s 2 + s + 3
5s +
1 + S6 +
4
s2
1
S2
12
s3
+
3
S3
P1 + P2 + P3
1 L1 L2 L3
The outputs of each integrator may be treated as state variables
X 1 ( s ) = 1s X 2 ( s )
X 2 ( s ) = 1s X 3 ( s )
X 3 ( s ) = 1s [3 X 1 ( s ) X 2 ( s ) 6 X 3 ( s ) + U ( s )]
Y ( s ) = 12 X 1 ( s ) + 4 X 2 ( s ) 5 X 3 ( s )
or
sX1 (s) = X 2 (s)
sX 2 (s) = X 3 (s)
sX 3 (s) = 3X1 (s) X 2 (s) 6X 3 (s) + U(s)
Y(s) = 12X1 (s) + 4 X 2 (s) 5X 3 (s)
dx1
dt
= x 2 (t)
dx 2
dt
= x 3 (t)
dx 3
dt
= 3x1 (t) x 2 (t) 6x 3 (t) + u(t)
y(t) = 12x1 (t) + 4 x 2 (t) 5x 3 (t)
let
x1
x = x 2
x 3
x&1
x& = x& 2
x& 3
A
B
6447
448
}
1
0 x1 0
x&1 0
x& = 0
0
1 x 2 + 0u
2
x& 3 3 1 6 x 3 1
C
6447
44
8 x1 D}
y = [ 12 4 5] x 2 + 0u
x3
Alternatively : Dual Phase Variable Form
sX 1 ( s ) = 6 X 1 ( s ) + X 2 ( s ) 5U ( s )
sX 2 ( s ) = X 1 ( s ) + X 3 ( s ) + 4U ( s )
sX 3 ( s ) = 3 X 1 ( s ) 12U ( s )
Y (s) = X 1 ( s)
6
x& = 1
3
y = [1 0
1 0
5
0 1 x + 4 u
12
0 0
0]x + 0u
In general
G(s) =
b1s n1 + ...+ bn1s + bn
s n + a1s n1 + ...+ an
Phase-Variable Form
0
0
:
A=
:
:
a n
1
0
.
a n 1
0
0
:
.
0
.
1
.. .. a1
0 ..
1 ..
0
:
B=
0
1
C = [bn
bn 1 .. b1 ]
D = [0]
Dual Phase-Variable Form
a1
a
2
:
A=
:
:
a n
1 0 .. 0
0 1 .. 0
.
:
.
0
. 1
0 .. .. 0
b1
b
B = 2
:
bn
C = [1 0 .. 0]
D = [0]
Multiple Inputs and Outputs
Single input, two-output system
T11 (s) =
5s2 + 4s 12
s 3 + 6s 2 + s + 3
T21 (s) =
3s2 + s 6
s3 + 6s2 + s + 3
Y ( s ) T ( s )
Y ( s ) = 1 = 11 U ( s )
T ( s )
Y2 ( s ) 1
42124
3
T (s)
T ( s) =
5s2 + 4 s 12
1
s 3 + 6 s 2 + s + 3 3s2 + s 6
1
0
0
0
0
1 x + 0u
x& = 0
3 1 6
1
0
12 4 5
y=
x
+
0 u
6 1 3
Two Inputs and One Output
T11 (s) =
5s2 + 4s 12
s3 + 6s2 + s + 3
10 s 2 + 7 s 8
T12 ( s ) = 3
s + 6s 2 + s + 3
6 1 0
5 10
7 u
x& = 1 0 1 x + 4
3 0 0
12 8
y = [1 0 0]x + [0 0]u
Transfer Function given State Space Form
x& = Ax + Bu
y = Cx + Du
(1)
(2)
(1) sX x(0) = AX + BU
set x(0) = 0
sX = AX + BU
( sI A) X = BU
1
X = ( sI A) BU
1
424
3
resolvent
matrix
Y(s) = CX(s) + DU(s)
(2) Y(s) = [C(sI A)1 B + D] U
144
42444
3
T (s) transfer function matrix
Example: Two Input Two Output
x&1 3 1 x1 4 6 u1
x& = 2 0 x + 5 0 u
2
2
2
y1 1 1 x1
y = 8 1 x
2
2
a b
1 d b
ad bc c a
c d
T ( s) = C ( sI A)1 B + D
1
1 1 s + 3 1 4 6
=
5 5 0
8 1 2
=
1 4 6
1 1 s
1
s 2 + 3s + 2 8 1 2 s + 3 5 0
6s
1 1 4s 5
1
8 1 5s 23 12
s + 3s + 2
s 9+s3+s18+ 2
= 27 s 63
s +3 s + 2
2
6 s +12
s 2 +3 s + 2
48 s 12
s 2 +3 s + 2
T ( s ) T12 ( s )
= 11
T21 ( s ) T22 ( s )
1
442443
T
+
U1(s)
T11
Y1(s)
+
T21
T12
+
U2(s)
T22
Y2(s)
+
Y = TU
Y1 T11 T12 U 1
Y = T
2 21 T22 U 2
Y1 = T11U1 + T12U 2
Y2 = T21U1 + T22U 2
State Transformation
x ( t ) = Pz( t )
where:
x& = Pz&
x& = Ax + Bu
Pz& = APz +Bu
1
1
z& = 1
P 2
AP
P2
Bu
3z + 1
3
A
x(t) is the old state
P is the non-singular matrix
z(t) is the new state
y = Cx + Du
y = CP
{ z+D
{u
C
New state space quadruple { A, B , C , D} for state z
where:
A = P 1 AP
B = P 1B
C = CP
D=D
Change of state does not change the input-output relationship
T ( s) = C ( sI A)1 B + D
so that
Proof
and
T (s) = C ( sI A )1 B + D
T(s) = T (s)
T (s) = CP ( sI P 1 AP )1 P 1B + D
= CP ( sP 1 IP P 1 AP ) 1 P 1 B + D
= CP[ P 1 ( sI A) P ]1 P 1 B + D
= CP P 1 ( sI A) 1 P P 1 B + D
using (AB)1 = B1 A1 if inverses exists
= C ( sI A) 1 B + D
= T (s)
Diagonalization of System Matrix
example
x&1 = x1 + x 2
1 1
x& =
x
x& 2 = x1 + x 2 or
1 1
y = x1 + x 2
y = [1 1]x
change of state:
let
1 1
x = Pz =
z
1
1
diagonal matrix
1
A = P AP
1 1 1 1 1 1 0 0
A = 12
1 1 1 1 1 1 0 2
z&1 = 0
z& 2 = 2 z 2
1 1
C = CP = [1 1]
= [0 2]
1 1
y = 2z2
solution is trivial
z1 (t) = z1 (0)
z2 (t) = e 2t z 2 (0)
where z(0) = P 1 x(0)
What transformation matrix P diagonalizes the systems matrix?
Ans.
P = [x1 : x 2 : ... x n ] where xi are the eigenvectors of A if A has distinct eigenvalues
example
2 1
A=
1 2
eigenvalues of A:
I A = 0
characteristic equation
+2
1
1
+2
( + 1)( + 3) = 0
eigenvalues
1 = 1
distinct diagonalizable
2 = 3
eigenvectors of A:
Ax i = i x i
for 1 = 1
x11
2 1 x11
1 2 x = 1 x
21
21
2x11 + x 21 = x11
x 21 = x11
1
choose x1 =
1
for 2 = 3
x12
2 1 x12
1 2 x = 3 x
22
22
1
therefore x 2 =
1
x 22 = x12
let x 22 = 1 x12 = 1
1 1
let P =
1 1
A = P 1 AP
1 1 2 1 1 1
=
1 1 1 2 1 1
1 1 1 3
= 12
1 1 1 3
2 0
= 12
0 6
1 0 1 0
=
=
0 3 0 2
Diagonalization Using Partial Fraction Expansion
Example
3
1 3
Y ( s)
s2 s 3
=
= + 4 + 4
U ( s ) s ( s 1)( s + 3) s s 1 s + 3
3
3
1
Y ( s) = U (s) + 4 U ( s) + 4 U ( s)
s
s 1
s+ 3
T (s) =
separate systems
z1
1
s
y1
+
z2
U(s)
1
s 1
+
Y(s)
y2
+
z3
1
s +3
y3
1
z1 ( s ) = U ( s ) z&1 = u
s
1
z 2 (s) =
U ( s ) z& 2 = z 2 + u
s 1
1
z 3 ( s) =
U ( s ) z& 3 = 3z 3 + u
s+3
y1 = 1z1
3
y 2 = z2
4
3
y 3 = z3
4
9
y = y1 + y 2 + y 3 = z1
z&1 0 0 0 z1 1
z& = 0 1 0 z + 1u
2
2
z& 3 0 0 3 z 3 1
3
3
z2 + z3
4
4
y = [1 34
z1
3
]z
4 2
z 3
Complete Conjugate Characteristic Roots
Example
2
4+ j
4 j
6s 2 + 26s + 8
=
+
+
T(s) =
2
(s + 2)(s + 2s + 10) s + 2 s + 1+ j3 s + 1 j3
0
0 x1 1
x&1 2
x& = 0 1 j 3
0 x 2 + 1u
2
x& 3 0
0
1 + j 3 x 3 1
x1
y = [ 2 4 + j 4 j ] x2
x3
Inconvenient to use complex numbers, so dont expand into complex factors
T(s) =
6s2 + 26s + 8
2
8s + 14
=
+ 2
2
(s + 2)(s + 2s + 10) s + 2 s + 2s + 10
0
0 z1 1
z&1 2
z& = 0
0
1 z 2 + 0u
2
z& 3 0 10 2 z 3 1
z1
y = [ 2 14 8] z 2
z 3
Example
T(s) =
5s + 1
6
8
8s
+
+ 2
+ 2
s 3 s + 4 s + 2s + 17 s 3s + 10
10
0
0
0
x1 3 0
x 0 4 0
0
0
2
0
1
0
d x3 0 0
=
dt x4 0 0 17 2 0
x5 0 0
0
0
0
0
0 10
x6 0 0
0 x1 1
0 x2 1
0 x3 0
+ u
0 x 4 1
1 x5 0
3 x6 1
x1
x
2
x
y = [6 8 1 5 0 7] 3
x4
x5
x6
Example
T(s) =
10s 2 + 51s + 56
3
7
3
=
+
+
2
(s + 4)(s + 2)
s + 4 s + 2 (s + 2) 2
JORDAN CANONICAL FORM
0 x1 1
x1 4 0
d
x 2 = 0 2 1 x 2 + 0 u
dt
x3 0
0 2 x3 1
x1
y = [3 3 7] x2
JORDAN BLOCK
x3
Time Response of Systems
x(t ) = e
A ( t t 0 )
x(t0 ) + e A(t ) Bu ( )d
t0
=e
A ( t t 0 )
x(t0 ) + A1[e A(t t ) I ]BU
0
where u(t)=U is a constant
where e At = 1{[ sI A] 1 }
3 1
2 0
Find eAt when A =
11
e At = 1{[ sI A]1}
1
s + 3 1
= {
}
s
2
1
s
1 s 2 +3 s + 2
2
s 2 +3 s + 2
1
1 s +1
2
s +1
+
+
s 2 +3 s + 2
s +3
s 2 +3 s + 2
2
s+2
2
s+2
1
s +1
2
s +1
e t + 2e 2t
=
t
2t
2e + 2e
1
s+2
1
s+2
e t e 2t
2e t e 2 t
Stability
x& = Ax
x(0) = x 0
System is asymptotically stable if all states approach zero with time - i.e. x(t)0 as t
This will happen if the eigenvalues have negative real parts.
Bounded-input, bounded-output (BIBO) stability means that the system output is bounded for
all bounded inputs. That is
u (t ) N <
y (t ) M <
In the absence of pole-zero cancellations, transfer function poles are identical to system
eigenvalues, hence BIBO stability and asymptotic stability are equivalent.
Example
T(s) =
Final Value Theorem :
lim y (t ) = lim sY ( s )
s 0
t
s 1
(s 1)(s + 2)
pole at s=-2
1
s 1
1
= lim
=
s1 (s 1)(s + 2)
s1 2s + 1
3
no pole at s=+1 since: lim T(s) = lim
s1
since only pole is at s=-2
system is BIBO stable since no pole in RHP or on imaginary axis
12
State-space realization
1 2
1
x& =
x + u
1 0
0
y = [1 1]x
Eigenvalues are 1 and -2 system is unstable in the asymptotic sense
Internal Stability
Based on transfer function description and is stronger than BIBO stability
D
R
General feedback system with disturbances
Internal stability requires that all signals within the feedback system remain bounded for all
bounded inputs
Requires nine transfer functions from inputs R,D,N to outputs U,V,W be stable
Sufficient if
1) 1+KGH has no zeros in RHP and on imaginary axis
2) KGH has no pole-zero cancellations in RHP or on imaginary axis
Controllability
A system is completely controllable if the system state x(tf) at time tf can be forced to take on
any desired value by applying a control input u(t) over a period of time from to until tf.
Observability
A system is completely observable if any initial state vector x(t0) can be reconstructed by
examining the system output y(t) over some period of time from t0 to tf.
Tests for Controllability and Observability
If the system matrix is diagonal then the tests are easy
13
e.g.
x1 1
d x2 0
=
dt x 3 0
x4 0
0 x1 1
0 x 2 1
+ u
0 x 3 0
4 x 4 0
y = [1 0 1 0]x
1, 2 , 3 , 4
e t ,e t ,e t ,e t
eigenvalues are
modes are
Modes e 3 t and e 4 t are uncontrollable since they are not connected to the control input
Modes e 2 t and e 4 t are unobservable since they are not connected to the output
e 1 t
e 2 t
e 3 t
e 4 t
mode
mode
mode
mode
controllable and observable
controllable but unobservable
uncontrollable but observable
uncontrollable and unobservable
For MIMO systems
-Uncontrollable modes correspond to zero rows of B
-Unobservable modes correspond to zero columns of C
Controllability Matrix
x& = Ax + Bu
M C = B M AB M .... M An 1 B
A system is completely controllable if and only if Mc has full rank
Example
4 1
A=
0 2
1
B=
0
M C = [B M AB ]
1 4
=
0 0
det(Mc)=0 not fully state controllable
14
Observability Matrix
C
CA
MO =
:
n 1
CA
A system is completely observable if and only if MO has full rank
Example
x&1 1 0 x1 1
x& = 1 1 x + 1u
2
2
y1 1 1 x1
y = 2 2 x
2
2
1 1
2 2
MO =
0 1
2
0
MO has two linearly independent rows (1 and 3) fully observable
MO MO
1 1
1 2 0 0 2 2
=
1 2 1 2 0 1
2
0
=
5 5
5 10
= 25 0
fully observable
15
