Escolar Documentos
Profissional Documentos
Cultura Documentos
L i n e a r C o n t r o l S y s t e m s
P r o f.
Darren
M. D a w s o n
N o t e s P r e p a r e d b y:
Siddharth Nagarkatti Aman Behal Pradeep Setlur
J X+ bX =T
Electrical Circuits L
+ e -
i C
di 1 + Ri + idt = e dt C dq i= q charge dt L
} }
Kirchoff's Voltage Law e : input voltage q : output charge i : current R : resistance L : inductance C : capacitance
L q + R q+
1 q=e C
Note : Two different systems are governed by similar looking Differential Equations
Control is the ''shaping" of the output of the Differential Equation by changing the input to the system via feedback
The Pendubot
An Example of a System that needs Control The Pendubot, short for PENDUlum RoBOT, is an under-actuated two-link mechanism. The system consists in two rigid links, the first of them actuated by a DC-motor and the second link is not actuated. The Pendubot is similar in spirit to the classical inverted pendulum on a cart or a rotational inverted pendulum. The Pendubot presents some unique features and challenges for control research. One can use the Pendubot to investigate linear control, nonlinear control, system identification, intelligent control, hybrid and switching control, and gain scheduling. Control Objective: To balance the pendubot and make the pendubot perform a specific acrobatic maneuver (i.e. the pendubot starts at the stationary vertical down position and must swing up to the inverted position)
Movie (.avi) of a working Pendubots [ECE 496 Senior Design Project Spring 1998]
Differential Equations
Definition A differential equation is an equation which involves differentials or derivatives For example:
2 d . y( t ) m x( t ) d t2 y
(n 1)
In this class, we will be concerned with ordinary differential equations of the form
an y ai
(n)
an bi
.....
a1 y
( 1)
a0 y bn y
(n)
(n
1)
.....
b1 u
(1)
b0 u
where u(t) and y(t) are functions of time. and are constants
(Note :
Time Invariant
The differential equations will be assumed to be time-invariant. That is, the coefficients, denoted by
ai and bi
Linearity
The differential equations are also assumed to be linear. What is a linear system ?
X1 H X1 X2
Y1
X2 H Y1 H Y2
Y2
are constants
Causality
we obtain an output which conforms to standard linear operations (add and subtract by a constant OR multiply and divide by a constant)
A system is called causal if the output depends only on the present and past values of the input. That is, if y(t) is the output, then x(t) depends only on the input less than or equal to t. A car, a light switch, a TV are all causal
x( )
n y an
d tn
Define
n 1 .d y 1 d tn 1 D
n
..........
d a1 . y dt
a0 .y u
d D ( ) dt
and
n ( )
d tn
a1 D .y
a0 y u
an D
1 n
....... D
n 1
a1 D ......
a0 .y u a1 D a0 0
( D) ( D)
Example
an
is called the
Find
( D) ( D) D ( D) ( D
so
y''+ 3 y' + 2y = x
x
Plant input
y
output
3 .D 2 0 1 ) .( D 2 ) 0 D 1 or D 2
Roots of
( D)
It is now obvious that this class involves the study of Differential Equations Where are we headed ? 1. Given x(t), it would be advantageous to have a general procedure for finding y(t) 2. It would also be advantageous to develop analysis techniques for examining systems without knowing the input 3. Develop the Laplace Transform Tool
} }
Problems
Solutions
Laplace Transforms
Why ? Because it is easy to work with an Algebraic Equation as opposed to a Differential Equation Laplace Transformations are used to change a differential equation into an algebraic equation
( f ( t ) ) F( s ) 0
where F(s) is the Laplace transform of p
f ( t ) .e
st
dt
}
and
f(t), s
denotes just to the right-hand side of zero (sometimes called zero-plus) When can you take a Laplace Transform ? t>0 and F(s) is convergent for some 0 , that is
Definition
( f ( t ) ) .e 0
p
0 t
dt <
then
f( t)
Example
( t) t
( t)
0 t
e 0
if p t
.e
0 t
dt 0
p
dt 1
1 0
.e
0 .t
0
p =
1 1 0
<
0> 0
or
0> 1
Region of Convergence
Note: We never worry about the ROC in Control Theory since we only use the one-sided Laplace Transform and all of our systems are causal t Example Find the Laplace Transform of e
0
p
t e .e
st
dt
1 s 1
(s
1 ) .t
0
p =
1 s 1
Note: We hardly ever use this method of calculating the Laplace Transform in practice; rather, we use the tables of Laplace Transforms and algebraic manipulation to find the Laplace Transform of a function
( f ( t ) ) F( s )
d f( t) dt t 0
sF ( s ) f ( ) d
f 0
2.
F( s ) s f 0
p
lim t 0
p
f( t) s
lim
sF ( s )
f( ) t
lim
f( t)
lim sF ( s ) s 0
5.
t f a ( f( t e
at .
a .F ( a .s )
sT
} } }
6. 7.
T) ) e f( t)
F( s ) a)
F( s
8.
f1 ( t ) * f2 ( t ) f1 ( t ) * f2 ( t )
F1 ( s ) .F2 ( s ) t
where
f1 ( ) .f2 ( t
) d
What are the above properties good for ? These properties can be used to manipulate a function into a form in which the transition, between the time domain and the frequency domain is eased (e.g. taking the Laplace Transform or the Inverse Laplace Transform)
d y dt sY ( s ) (s
4.y 2 e y 0
p
3t
y 0
3 2 s
p
4 Y( s ) 2 s 2 3
} } } } } }
Turning a DE into an AE
4 ) Y( s )
y 0
Y( s )
(s 2 s 2 s
3) . ( s 2 3 3 s 1 s
3t
4)
y 0 s 4 3
Solve the AE
Y( s ) Y( s )
4 4 e
4t
y( t )
2e
The answer
y( 0 ) 3
3t 4t
y( t ) 2 e e d 3. t y 4.y 2.e dt d 3t 4t 2e e dt 6e
3t
4 2e 8e
3t
e 4e 2e
4t
2e 2e 2e
3t
4e
4t
3t
4t 3t
3t 3t
} } }
u
System
an
(n
1)
.....
a1 y
( 1)
a0 y bn
(n
1)
.....
b1 u
(1)
b0 u
If we assume that all initial conditions are zero, then we can take the Laplace Transform
s
so
an Y( s )
. sn
......
1 . sn 1
a1 . s
1 1
a0 . Y ( s ) ..... ...... b1 . s a1
.s
bn b0 a0
. sn
.....
b1 . s
b0 . U( s )
bn s
n
an
. sn
. U( s )
U( s )
y( t )
Y( s ) H( s ) U( s )
bn s
n
. sn 1
1 1
..... ......
b1 . s a1 . s
b0 a0
an
. sn
} }
Remember : Set the initial conditions to zero when calculating the transfer function
10
H( s )
s4
8s s
3
25 s 6s
2
31 s 6
15
11 s
Step 1: Make the transfer function strictly proper,i.e,, order of (denominator) > order of (numerator) Long Division
s s
3
2
4
6s
11 s
8s 25 s 31 s 3 2 . . 6s 11 s 6 .s 2s 3 2s
3
15
Dont let yourself get burned on the test
14 s 2 12 s 2s
2
25 s 22 s 3
15 12
Stop when the order of the numerator is one less than the denominator
3s
H( s ) s
2 2s
2
H1 ( s ) 3s
2
H1 ( s )
3 6
6s
11 s
Step 2:
Find
h1 ( t ) s
3
H1 ( s )
2s
2 2
3s
3 6 (s K2 s 2 s
6s
H1 ( s )
Now Find
11 s K1 s 1
2s 3s 3 . 1 ) ( s 2 ) .( s 3 ) K3 3
K1 K2
and
K3
11
Residue Formula:
Ki
si .H( s )
s si
2
K1 ( s
1 ) H1 ( s )
1
=
2s 3s 3 . ( s 2) ( s 3) 2s 3s 3 . ( s 1) ( s 3) 2s 3s 3 ( s 1 ) .( s 2 ) 1
2 2
1
=
2 3 3 1 ( 1 ) .( 2 ) 8 6 3 ( 1 ) .( 1 ) 5
K2 ( s K3 ( s H1 ( s )
2 ) H1 ( s ) 3 ) H1 ( s ) 6 (s 3)
2
2
=
2
=
s 5 2)
18 9 3 6 ( 1 ) .( 2 )
(s
(s
1)
(s
2s 3s 3 . 1 ) ( s 2 ) .( s 3 ) 5 .exp ( 2 .t ) 2 ( t) h1 ( t ) exp ( t )
After taking the Inverse Laplace Transform (see the table), we have
h1 ( t ) 6 .exp ( 3 .t ) 1
So
answer The whole point of the PFE is to get the expression in the frequency domain into a form we can recognize in the tables
h( t ) =
( s)
H( s ) s
bn
n
.s n
1
1 n 1
..... ......
b1 .s a1 .s A1 s sn
b0 a0
an
.s
sn ..
}
A2 sn
.....
H( s )
K1 s K1 s1
.....
....
Kr s
with
Aq sn
q
sr
s A1
Find
Kr
Find
Aq
12
Aq
q sn .H( s )
s sn sn
q.
Aq
1 d 1! ds 1 d2 2! d s2
H( s )
s sn
and so on
Aq
sn
q.
H( s )
s sn
H( s )
11 s
23 s 1)
2
s( s K1 ( s .H( s ) ) A2 A1 (s
1 K1 s
A1 s 1 (s
A2 1)
2
one distinct root (s=0) one repeated root (s=-1) multiplicity two (q=2)
s 0
1 1 1
=
2 1 ) .H( s ) 2 1 ) .H( s )
11
1
=
23 1
2
11 23 s s 1 s 1
d (s ds
=
s s ( 22 s 23 )
1
2
d 11 s ds 23 s
11 s s
2
1 s 1
= 10
H( s )
1 s
10 s 1 (s
11 1)
2
h( t ) 1
10 .exp ( t )
11 .t .exp ( t )
answer
13
d tn
Property 10.
x ( t ) s X( s )
x 0
(n
2)
( 1)
....
(n
1)
( t ) .e
st
dt 1
note
( ( t) ) 1
} ( t)
f ( t ) . ( t
T) dt f ( T)
Transfer Function
Input / Output approach to system modeling
u( t ) H( s ) Y( s ) H( s ) U( s ) y
n
y( t )
an
..... .s n
1
a1 y
1
( 1)
a0 y bn b1 .s a1 .s
(n
1)
.....
b1 u
( 1)
b0 u
Y( s ) H( s ) U( s )
bn s
n
.....
1
b0 a0
an
.s n
......
Roots of the numerator - zeros of H(s) Roots of the denominator - poles of H(s)
Stability
14
Example
+
Vc
Given +
Vc
1F i y 1
_
u
_
Vc 0 y i
1 d V dt c
=>
{ { {
u Vc
y d y dt
d V dt c
d d u V dt dt c d u y dt d y dt
Circuit Analysis
Substitute for
d V dt c
( u( t ) ) sY ( s ) Vc 0
p
2 s 1
sU ( s ) y 0
p
u 0 u 0
p
y 0 1
Y( s )
Calculate y statement
}
1
u 0
2)
sU ( s ) Y( s ) (s Y( s ) (s
Use Partial Fraction Tool
2 sY ( s ) 2s 1) 2s 1) 2 (s 1)
2 2
1 2 1 1 1 B s 1
Y( s ) s 1
s s
} }
Simplify
}
1 s
Y( s )
}
1
d (s ds
2 1) .
2s (s 1)
t 2
s
t
}
t
y( t )
2 te
2e
1e
2 .t .e
How does a Mathematician check this answer ? How does an Electrical Engineer check this answer ?
Answer
15
More Formulas
New Laplace Transformation Table Entries at
sin 0 t cos 0 t
0 (s a)
2
0 a 0
and
at
s (s a)
2
Y( s ) u( t ) ( t ) H( s )
u( t ) ( t )
( u( t ) ) 1
note
Y( s ) H( s ) .U( s ) H( s ) .1
hence
( u( t ) ) 1
y( t ) h ( t )
when
u( t ) ( t )
d
2
2 y
d t2 s Y( s ) Y( s )
2
2.
d y dt
5 .y u ( t )
y( 0 )
d y ( 0) 0 dt
2 s Y( s ) 1
5 Y( s ) U( s ) U( s ) 1 h( t ) = 5 ( H( s ) )
s 2s 5 Y( s ) 1 H( s ) 2 U( s ) s 2s
16
Example - Continued
Now let
u( t ) 3
find
y( t ) 3 U( s ) s 3 s s
2
Y ( s ) H ( s ) .U ( s )
2s
The roots of the polynomial in paranthesis are given by A partial fraction expansion is given by
j2
and
j2
Y( s )
A0 s s
B0 1 j2 s
C0 1 j2
There are no complex numbers in the tables, so we use a different PFE method Another partial fraction expansion is given by
Y( s )
A s
Bs s
2
C 5
2s
1. Since there are 3 roots, you need 3 weighting coefficients 2. The degree of the numerator polynomial must be one less than that of the denominator
PFE Rules
Y( s )
3 .1 Bs C 3 5 s s2 2 s 5 s s2 2 s 3 2 6 2 s s 3 Bs Cs 3 5 5 B
and
}
}
"A" was calculated with the Residue Formula "B" and "C" are found with the common denominator method
3 5 3 .1 Y( s ) 5 s s 2 s
2
6 5
3 s 2 5 s2 2 s 5 s 2 s 1)
2
}
1 1)
2
1 2
2
s (s 1)
1
2
2s
5 (s
(s
1 2 (s
2 1)
2
s a)
2
a 0
2
0 (s a)
2
17
Y( s )
3 .1 5 s 3 5
3 s 1 5 ( s 1) 2 22
3 10 ( s
2 1)
2
Final Form
y( t )
3. exp ( t ) .cos ( 2 .t ) 5
3 . exp ( t ) .sin ( 2 .t ) 10
Impulse Response Step Response 0.7 0.6 0.5 Amplitude 0.4 0.3 0.2 0.1 0 0 2 4 Time (sec.) 6 8 10
Plot of y(t)
The overshoot and slight ripple is cause by the complex poles which also give rise to cos(t) and sin(t) terms
y( 0 )
d y dt
5 .y u ( t )
u( t ) 3
18
D( s ) s n an
1
an
.......
a1 s
a0
D1 ( s ) s
.s n
1
.......
3
D2 ( s ) an D1 ( s ) D2 ( s )
.s n
an
.s n 1
.......
Formulate
D1 ( s )
and
D2 ( s )
h1 s h2 s h3 s
1 1 h4 s 1 ...hn s
If all
( Note OLHP - open left half plane-poles have negative real parts)
D( s )
3
stable ?
D( s ) s D1 ( s ) s
3
6s
12 s
12 s
D2 ( s ) 6 s
Formulate
D1 ( s )
and
D2 ( s )
D1 ( s ) s 3 12 s D2 ( s ) 6 s 2 8
1 s 6
=>
6s
s s
3 3
12 s 8 s 6 32 s 3
D1 ( s ) 1 s D2 ( s ) 6
1 6s 8 32 . s 3
2
Continue Dividing
19
9 s 16 D1 ( s ) 1 s D2 ( s ) 6 1 6s 8 32 s 3
2 =>
32 s 3
6s
8
2
6s
D1 ( s ) 1 s D2 ( s ) 6
1 9 s 16 1 4 s 3
Algebraic Simplication
h1
1 9 h2 and 6 16
h3
4 3
have all negative real parts (i.e., it is stable)
D( s )
20
( s) s D1 ( s )
3 3
14 s s
56 s
K 14 s
2
1 s 14 K s s
3 3
Long Division
56 s
2
D2 ( s ) 14 s
56 s K. s 14 56 K s 14
D1 ( s )
1 s D2 ( s ) 14
56 14 s
2
K s 14 K
s 14
1 14 s 56
2
K K s 14
Performing long division again
D1 ( s )
s D2 ( s ) 14
1 s 56 14 K 14 14 56 K 14 s 56 K 1 K 14 1 56 K K< 784 K 14 s 56 K 14
s. 56
2
14 K 14
14 s K 2 14 s K
h1
1 14
h2
h3
For each
hi> 0 >0
we must have
14 56 K 14
therefore
56
=>
K> 784
=>
21
d t2 U
2.
Y n
2
2 . . n .s . n
The poles of
H( s )
are
s s
j . n . 1
and
and
. n
j . n . 1
j . d
j . d
d j
j .
j . n j . d n - length of vector j . d j . n
. n d n . 1
2
s - plane
{
s
cos
axis
j . d
If If If
poles are negative and real poles are real and equal poles are complex conjugate pairs poles are imaginary
. n s j . n
j . n . 1
and
j n
and
. n
j . n . 1
If
determines how much cosine and sine action an output might have
22
If
t .e
n .t
1
=
n n
} }
If
sin n .t
If
0< < 1 1
n
2
.e
. n .t
.sin 1
t s
2.
n
2
2 . . n .s
If
> 1
roots are distinct and real; hence standard PFE/Tables can be applied
23
which means that all poles are in the OLHP (Note OLHP - open left hand plane - poles have negative real parts)
Examples
Transform Pairs
{ {
H( s ) s
1
2
s-plane
j 1
Not Stable
j
s-plane
h ( t ) sin ( t ) 1 1 ) .( s 2 2) 1
Stable
Transform Pairs
H( s )
Pole Plots
(s
h ( t ) exp ( t )
exp ( 2 .t )
s-plane
Transform Pairs
H( s ) h( t )
(s 1. exp ( t ) 2
1 1 ) .( s
1) 1. exp ( t ) 2
Not Stable
24
25
U(s)
H(s)
Y(s)
Summer
X(s)
+ -
Y ( s) Z(s)
U(s) Y ( s) X ( s) multiplier U ( s) Z ( s)
} Mathematical Expression
Y ( t) Y ( t) K.X ( t)
X ( t)
Mathematical Expression
K X ( s) Y ( s) Y ( s ) K.X ( s )
Mathematical Expression
26
Canonical Form
R ( s) + H ( s) G(s) Forward Transfer Function H(s) Feedback Transfer Function L ( s) H ( s) .Y ( s) R ( s) L(s) G ( s) Y ( s)
} } } } }
from Block Diagram from Block Diagram substitute for L(s) algebraic manipulation answer
Y ( s) L ( s) .G ( s) Y ( s) ( H ( s) .Y ( s) Y ( s) ( 1 Y ( s) R ( s) 1 R ( s) ) G ( s)
H ( s) G ( s) ) R ( s) .G ( s) G ( s) G ( s) .H ( s)
R ( s)
G ( s) G ( s) .H ( s)
Y ( s)
The need for simplification of Block Diagrams spawned the creation of the rules and procedures
27
simplified
Parallel X
P1
P2 . X P1 P2 + + Y
P1
P2
simplified
}
Y
P1 Y P2 . P2 Feedback loop Y P1 X P2 . Y P1
P2 . X
+ -
P1 P2
X 1
P1 P1 . P2
simplified
X 1 P2
+ -
Y P1 . P2
Y=
equivalent to above block.
1 P 1P 2 X P2 1 + P 1P 2
28
+ + 1 P
Z Y
Z P. X
X Y +
X Y
P P
+ +
Z P( X
Y)
X P
Y P. X
X X
X X
P 1/P
Y P. X
Bottom Line : If you move wires or blocks, you must ensure the mathematical expression remains equal to the same thing that you started with
29
Multiple Inputs
Step 1: Step 2: Step 3: Step 4: Example
When we have multiple inputs, each is treated independently of the other inputs. Set all inputs except one input equal to zero Calculate the response due to the non-zero input. Repeat step 1 and step 2 as needed Add all the responses together. U R + G1 + + G2 C
Find C(s)
Set U = 0 R + G1 . G2 CR
}
Cu
CR = Set R = 0
G1 . G2 1 G1 . G2
.R u +
}
G2 G2
G1
} }
+ + 1
Cu
G1 G2 1 G1 . G2
}
Block Diagram Simplification Rules G2 G1 . G2 U
G1 . G2 1 G1 . G2
.R
30
(Method 1)
Example 1
G3 + R + + + G1 G4 G2 + C
H1
}
G2 G3 C
Find C ( s) R ( s)
H2
+ -
+ +
G1 G4
H1
H2
31
+ 1
G1 G4 H1 G1 G4
G2
G3
}
H2 R 1 G1 G4 G2 G3 1 H1 G1 G4 G1 G4 H2 1 G2 G3 H1 G1 G4 C
32
Example 2
R + + K 1 s 1 C
1 s 1 s 1 1
}
1 s 1 C
Find C ( s) R ( s)
0.1
R -
+ K
+ + 0.1
1 s 1 s 1 1
}
1 s 1
+ -
H ( s)
} } } }
add parallel paths close feedback loop
H ( s)
s s 1 s
.1
( s .1 ) 1 ( s 1)
C ( s) R ( s)
K . 1 KH ( s ) s 1
C ( s) R ( s)
= 1
K . 1 K( s 0.1 ) s 1 s 1
answer
33
34
m Y Y m .x b
mathematical expression
Y Y 3 .X
X
4
Z Z
20
( 4 .X)
}
X Z
mathematical expression
10
200
Y Y 10 .X
Z Z ( 20 .Y)
Z ( 200 .X)
35
A32
A33 X3 A43 X4
X1 to X2 to X3 to X4 X2 to X3 to X4 X2 to X3 to X2
} }
Examples of Paths
36
Definition - an input node is a node with only outgoing branches Definition - an output node is a node with only incoming branches
Example Example
Definition - a forward path is a path from input node to output node Example Definition - a path gain is the product of the branch gains encountered Example --
X1 X4 X1 , X2 , X4
Definition - A loop gain is the product of the branch gains of the loop Example -loop gain for a loop
R
+
C
Block Diagram
C
Signal Flow Graph
37
SFG
C P =H = i i R i Pi Pjk
= ith forward path
Masons Rule
= 1 (1) K +1 PJK
K J
} Original Formula
= 1 - (sum of all loop gains) + (sum of gain-products of 2 non-touching loops) - (sum of gain-products of 3 non-touching loops) + (....... 4.........) - ......
} Alternate Formula
Pi
being eliminated
Definitions - Two loops, paths, or a loop and a path are non-touching if they have no nodes in common
38
H
1) Find forward paths
P1 G
2) Find loops
R
One forward path
P11
GH i s GH
One loop
3) Find and
}
}
= 1 - (sum of all loop gains) + (sum of gain-products of 2 non-touching loops) - (sum of gain-products of 3 non-touching loops) + (....... 4.........) - ......
0 1
recalculate with
P1
removed
H
4) Calculate H(s)
H( s )
P1 1
1
G GH
Does this agree with what you know the answer is supposed to be ?
39
Example
G3
+ + +
R
+
G1
G4
G2
C
+ worked this example before with standard block diagram simplification rules
H1 H2
1) First, draw the Signal Flow Graph
}
1 1
G3
G1 G4 G2 H1 H2
1
P1 G1 G2 G4
1
G1 G4
G2
P2 G1 G3 G4
3) Find the loops
G1 G4 G1 G4
G3
P11 G1 G4 H1 H1 P12 G1 G2 G4 H2 G1 G4 G2
1
H2
40
P21
G1 G3 G4 H2
G1 G4
G3
4) Find
and
i s P12 P21
0
H2
P11
i 1
Pi
being eliminated
Recalculate with
P1
removed
G3 H1
Recalculate with
P2
removed
H2
1
H1
5) Find H(s)
H2
P11 + P2 2
H =
H= 1
G1 G4 G2 ( 1 ) G1 G4 H1
G1 G3 G4 ( 1 )
G1 G2 G4 H2
G1 G3 G4 H2
41
G2
+
R
+ +
G1 H1
G2 G1
1 1
R
1 1
H1 C
1 1
P1 G1
R G1 G2
1 1
P2 G2 R
1 3) Find the loops 1
C G1
P11 G1 H1 H1
4) Find and
i s
= 1 G1H1
= 1 - (sum of all loop gains) + (sum of gain-products of 2 non-touching loops) - (sum of gain-products of 3 non-touching loops) + (....... 4.........) - ......
1 = 1 }
Recalculate with
P1
removed
G2
H1
42
2 = 1 G1H1
Recalculate with 1
P2
removed
G1
H1
5)
H (s) =
P G + G2 (1 G1 H 1 ) 1 1 + P 2 2 = 1 1 G1 H1
R( s )
+ -
1/s
1/s 1 1
C( s )
1 1) Draw SFG
1/s 1 1
1/s
P1
R
1 1/s 1 1 1/s 1 1
1/s
s
1
43
1/s
s
1
1 1 1 1/s 1
s i s
4) Find and
= 1 ( A + B + C ) + AB
= 1 - (sum of all loop gains) + (sum of gain-products of 2 non-touching loops) - (sum of gain-products of 3 non-touching loops) + (....... 4.........) - ...... 1
1 = 1
5) Find
Recalculate with
P1
removed
C( s ) H( s ) R( s )
1 1 1 = 1 1
C P = 1 1 R
C R s2 B C)
1
= 1
s2
3 1
(A
AB
s2
C( s )
s .( s 2
5 2 1
1) 3 2 1 5
s( s
a) ( s
b)
R( s )
a C( s )
b ab s
1 +
1 +
a ( a b) s a
+ 1
b ( b a) s b
at + 1
}
e
bt
PFE
c( t)
ab
(b
a) a
(a
b) b
44
b1 u
1 1/s 1/s 1/s
b2 b3
a2 a1 a3
P1
b3 s3
u
1 1/s 1/s 1/s
y b3
P2
b2 s
2
b2 u
1 1/s 1/s
P3
b1 s u
1 1/s
b1 y
a1 s
1/s
a1
45
1/s
1/s
a2 s2 a2
a3 s3 i s C)
1/s
1/s
1/s
3) Find
and
a3
(A
= 1 - (sum of all loop gains) + (sum of gain-products of 2 non-touching loops) - (sum of gain-products of 3 non-touching loops) + (....... 4.........) - ...... removed
} Recalculate with
P1
b1 b2 a2
a1
} Recalculate with
P2
removed
a3
b1
1/s
a1
a2 a3 P3
removed
b3
3 = 1
} Recalculate with
b2 a1 a2 a3
1/s
b3
46
4) Find H(s)
Y (s) P = i i = U (s)
P1
1
P2 B
P3 C)
(A
b3
= 1 s3
b2
s2
b1
s
a1 s
a2 s2 b2 s
2
a3 s3 b3 a3
b1 s 2 s
3
a1 s
a2 s
Example
+ -
G1 H1
+ -
G2 H2
G1
G2
R H1
1) Find the forward paths
H2
P1 G1 G2
2) Find the loops
R
1
Y G1 G1
1
G2
P11
G1 H1 H1
47
P12
G2 H2
G2
H2
3) Find and
i s
= 1 (P 11 + P 12 ) + ( P 11P 12 )
= 1 - (sum of all loop gains) + (sum of gain-products of 2 non-touching loops) - (sum of gain-products of 3 non-touching loops) + (....... 4.........) - ......
Recalculate with
P1
removed
H1
4) Find H(s)
H2
Y( s ) R( s )
P1 1
G1 G2
= 1
G1 H1
G2 H2
G1 G2 H1 H2
R( s )
1
G1 G1 H1
1
G2 G2 H2
Y( s )
Y( s ) R( s )
= 1
G1 G2 G1 H1 G2 H2 G1 G2 H1 H2
48
49
R( s )
B
Gp ( s )
K .Gp ( s ) Y( s ) R ( s ) 1 K .Gp ( s ) s Gp ( s )
m n
}
......
am bn
1 1
.s m
1
.s n
....... n
a0 N ( s ) b D( s )
0
N( s )
Y( s ) R( s )
K. 1
N( s ) D( s ) K .N( s ) N ( s ) D ( s ) K .N ( s ) K. D( s ) K .N ( s ) 0 K
is varied from 0 to Root Locus
The closed loop poles are the roots of the characteristic equation
( s ) D( s )
The location of the roots of
( s)
( s) 1
K.
N( s ) D( s ) D( s )
K .N ( s ) 0
50
i =
The root loci may be found on portions of the real axis to the left of an odd number of open-loop poles and zeros. The root loci are asymptotic to straight lines, for large values of s, with angles given by
=
Let RD (relative degree) = n - m
(1 + 2k ) nm
k = 0,1,..., n m 1 n = no. of finite open loop poles m = no. of finite open loop zeros
120
90
72
RD 1
RD 2
RD 3 x
RD 4
in the asymptote charts above
RD 5
51
Example
Plant
R( s )
B
Gp ( s )
Y( s )
Gp ( s )
1 s .( s 2 ) .( s 4 ) 1 K. Y( s ) s .( s 2 ) .( s 4 ) R( s ) 1 1 K. s .( s 2 ) .( s 4 ) ( s) 1 K. s .( s 1 2 ) .( s
s 0, 2, 4
4)
j j
; ;
-4 -2
; ;
-4 -2
centroid
centroid RD 3
2 3
-4
; ;
-2
Imag Axis
a. Locus must be symmetric about real axis b. Open loop zeros at infinity = 3 c. Locus starts at open loop poles and goes to the open loop zeros
-2
Note: Locus is symmetric about the real axis because complex poles always come in conjugate pairs
-4
-6 -6
-5
-4
-3
-2
-1
Real Axis
52
Example
Plant
R( s )
B
Gp ( s )
Y( s )
Gp ( s )
(s K. 1
1 ) .( s
s 2 3 3j ) .( s
3j )
s = 2 s = 1,3 j3
Y( s ) R( s )
(s (s
1 ) .( s
K.
s 2 3 3j ) .( s 3 3j ) s 2 . 1 ) ( s 3 3j ) .( s 3 3j ) s 2 3 3j ) .( s
( s) 1
K.
(s
1 ) .( s
3j )
;
|
-3 -2 -1
;
|
centroid
R ;
R ;
-3 -2 -1
;
5 2
Centroid
1 2
( 2)
R ;
-3 -2 -1
RD 2
10 8 6
4 2 0 -2 -4 -6 -8
a. Locus must be symmetric about real axis b. Open loop zeros at infinity = 2 c. Locus starts at open loop poles and goes to the open loop zeros
-1 0 -4
-3
-2
-1
R ea l A xis
53
Example
R( s )
B
Gp ( s )
Y( s )
Gp ( s )
(s K.
1 1 ) .( s
3)
1, 3
Y( s ) R( s )
1 ( s 1 ) .( s 3 ) 1 K. ( s 1 ) .( s 3 ) 1 1 ) .( s
}
}
( s) 1
K.
(s
3)
j
-3
j
centroid
; ;
-1
-3
; ;
-1
Centroid
3 2
; ;
-3 -1
2.5
RD 2
Imag Axis
2 1.5 1 0.5 0
Step 4. Draw the Root Locus a. Locus must be symmetric about real axis b. Open loop zeros at infinity = 2 c. Locus starts at open loop poles and goes to the open loop zeros
Real Axis
54
Example - Continued
Open Loop System
R( s ) (s 1 1 ) .( s Gp ( s ) 3)
S te p R e s p o ns e
Y( s )
Gp ( s )
Let R(s) = Unit step
0 .3
Y( s ) 1 3
1 3 s 1 e 2
Amplitude
1 2 s
t
1 6 1 1 e 6 s
3t
0 .2 5
0 .2
0 .1 5
0 .1
0 .0 5
y( t )
Y( s )
B
R( s )
Let R(s) = unit step
Gp ( s )
K 30 K 10 K 4 K 1
0.8 Amplitude
0.6
}
10
0.4
0.2
0 0 2 4 Time (sec.) 6 8
Note: Higher values of K speed up the closed-loop response when compared to the open-loop response
55
Example
R( s )
Gp ( s )
Y( s )
Gp ( s )
s .( s K.
3 ) .( s
1 j .7.4 ) .( s
j .7.4 )
Y( s ) R( s )
( s) 1
K.
s .( s
3 ) .( s
j .7.4 )
; ;
-3 ;
-3
; ;
centroid
Centroid
3 4
9 4
2.25
;
-3
; ;
RD 4
56
Step 4. Draw the Root Locus a. Locus must be symmetric about real axis b. Open loop zeros at infinity = 4 c. Locus starts at open loop poles and goes to the open loop zeros
8 6 4
Imag Axis
2 0
-2 -4 -6 -8 -4
-3
-2
-1
Real Axis
57
R( s )
B
Y( s ) K Gp ( s )
( s) 1 ( s)
K .Gp ( s ) s clp
characteristic equation
=0
Magnitude Condition
( s) 1
K .Gp ( s ) K .Gp ( s )
s clp
=0
=>
K .Gp ( s )
= -1
s clp
s clp
} }
K . Gp ( s )
s clp
=1
Magnitude Condition
Angle Condition
K .Gp ( s )
= -1
}
=
s clp Angle ( 1 )
Angle K .Gp ( s )
s clp
Angle Gp ( s )
= -180
s clp
Use -180 degrees as opposed to 180 degrees by convention (only here) (Note K is a real number)
Angle Gp ( s )
= -180
Angle Condition
s clp
58
Example
R( s ) Gp ( s ) (s s 2 1 ) .( s 4)
B
Y( s ) K Gp ( s )
Y( s ) R( s )
K. 1
s 2 ( s 1 ) .( s 4 ) s 2 K. ( s 1 ) .( s 4 )
closed loop TF
1 0.8 0.6
Imag Axis
0.4 0.2 0
Magnitude Condition
K . Gp ( s )
K. K.
5
=
(s
s 2 1 ) .( s
4) 1
=1
s
=>
5 K 4 3
3 . ( 4) ( 1)
Angle Condition
Angle Gp ( s )
=
Angle ( s
2)
Angle ( s
4)
Angle ( s
1)
= - 180
Angle ( 3 )
( Angle ( 1 ) )
59
d t2
where
( t)
Ra La
B .d ( t) J dt ( t)
Ra B . La J
Ga .If La .J
. ( t )
.V ( t ) a
Va ( t )
d t2
( t)
6. 1
d ( t) dt
5 . ( t ) Va ( t )
Plant Model
( s) Va ( s ) s 2 ( s) s
Poles are at 2
6 .s 6 .s
5 5 (s 5
Gp ( s ) Va ( t ) 1 ) .( s 5) 0 s
Gp ( s ) 1
2
( t) 5
6 .s
1 and s
then
Let
Va ( t ) ( t )
( t)
1
=
Gp ( s ) .Va ( s )
Pole Plot
4
( s) ( t)
Imag Axis
3 2 1 0
-1 -2 -3 -4 -6
-5
-4
-3
-2
-1
Real Axis
Definition
Dominant Poles are the closest poles to the
The dominant pole corresponds to the slowest mode The open loop systems output response will be dominated by e t Why ?
60
Va ( s ) ( s) 1 1 ) .( s 1. exp ( t ) 4
1 s
hence
( s ) Gp ( s ) . 1 . 1 20 s 5
1 s
s( s 1 5
1 .1 5) 5 s
1. 1 4 ( s 1)
}
0.2 0.18 0.16 0.14
PFE Tool
Step Response
( t)
1. exp ( 5 .t ) } 20
Problem: The response of the system maybe too slow to suit our specifications Solution: Add Feedback
Va ( s ) R( s )
B
( s) Gp ( s )
( s) R( s ) 1
Find K so that the closed loop poles are at
K .Gp ( s ) K .Gp ( s )
3 1 1 ) .( s
j .3
and
j .3
4
K. K.
(s
5)
s j .3 )
3 1
j .3
=1
Imag Axis
3 2 1 0
( 2
1 j .3 ) .( 2
-1
K 13 s 3
-2 -3 -4 -6
j .3
and
j .3
-5
-4
-3
-2
-1
Real Axis
61
( s) R( s ) 1 ( s)
K .Gp ( s )
R( s ) s 2 ( s) s. s
13 ( s 1 ) .( s 5 ) K .Gp ( s ) 13 1 ( s 1 ) .( s 5 ) 13 1 R( s ) . 6 s 18 s 13
6s
18 1 s 18 1 3
2
( s ) 13 .
1 .1 18 s
(s
3)
13 .
1 .1 18 s
1. s 3 3 18 ( s 3 ) 2 3 2
( s ) 13 . ( t) 13 18
1 .1 18 s
1 . s 3 18 ( s 3 ) 2 3 2
3 (s 3)
2
13 . exp ( 3 .t ) .sin ( 3 .t ) 18 e
3t
13 . exp ( 3 .t ) .cos ( 3 .t ) 18 e
t
Closed Loop
Open Loop
Open Loop
Amplitude
Note that the closed loop response is faster than the open loop response
62
Compensation
Gp ( s )
Let
R( s )
B
Compensator
Plant
Gc ( s )
Gp ( s )
Y( s )
1 s .( s 1)
0.5
0.5
K Y( s ) s .( s 1 ) K 2 R( s ) K s s 1 . s ( s 1)
Root Locus tells us that desired closed loop poles are possible with
1.5
Gc ( s ) K
Imag Axis
0.5
0.5
j .0.5
-0.5
0.5
j .0.5
}
1
-1
-1.5 -2
-1.5
-1
-0.5
0.5
Real Axis
( s) s
K 0.5
Actual denominator
d ( s ) ( s
j .0.5 ) .( s
0.5
j .0.5 ) s K 0.5
0.5
Desired denominator
63
( s) 1
K s .( s 1 )
0.5
j .0.5 K 1 4 1. 1 4 4 1 4 1 2
=0
s . s
1 s 0.5
j .0.5
hence
1 s .( s
4
y( t ) 1)
Y( s ) Gp ( s ) .R ( s ) 1 1)
Amplitude 3.5 3 2.5 2 1.5 1 0.5 0 0 1
Step Response
y( t ) A
Bt
C .e
}
2 Time (sec.) 3 4 5
y( t )
lim s
s .F ( s )
y 0
lim s
s
2 s .( s
0 1)
} }
1)
64
R( s ) R( s ) 1 s 0.5 s .( s
1 1)
Y( s )
closed loop TF
Y( s )
s
0.5 .t .
( 0.5 )
}
}
y( t ) A
Be
cos ( 0.5 .t )
0.5 .t .
sin ( 0.5 .t )
FVT :
y( ) s lim 0 s . s2
Step Response
1 0.9 0.8
IVT :
y( 0 ) s lim
2 s. s
0 0.5
} y( t )
0 5 10 Time (sec.) 15 20
65
2 . n .s
1. Percent Overshoot (PO) max value of output - final value of output PO = final value of output
100 90 80
PO 100 .exp 1
.
2
Graphical Solution
delta
2. Settling Time - Time required for the system output to remain within b% of the final value Typical values:
b 2%
ts
4 . rad
and n
b 5%
ts
3 .
3. Rise Time - Time for the system output to go from 10% to 90% of the final value.
tr
where 2
n . 1
rad
is in radians
cos ( )
4. Peak Time - Time required for system output to reach first peak of the overshoot.
tp
n . 1
2
Above relationships are used to convert time domain specification to pole locations.
66
1.4
Step Response
PO
1.2
1.02
Amplitude
1 0.8
0.98
0.6
0.4
0.2
ts
0 0
tp tr
10
15
20
Time (sec)
tr
rad
2
ts
n . 1 tp n . 1
4 .
}
. 1
2
Conversion formulas
PO 100 .exp
n
s-plane
j d cos
tr , ts , tp , PO
} pole locations
closed loop pole locations
Conversion Formulas
67
R( s )
-
Gp ( s )
Y( s )
e( t) r( t)
y( t )
SSE t
lim
r(t)
e( t)
lim sE ( s ) s 0
} SSE
y(t)
Amplitude
Time (sec)
E( s ) R( s )
Y( s ) .R ( s )
K .Gp ( s ) 1 K .Gp ( s )
.R ( s )
E( s ) R( s )
Substitute Y(s)
R( s ) . 1 E( s ) K , Gp ( s ) 1
K .Gp ( s ) .R ( s )
K .Gp ( s ) R( s )
Common denominator
If
and
E( s )
1 .R ( s ) K .Gp ( s )
Simplify
SSE
lim s . 1 s 0
1 .R ( s ) K .Gp ( s )
68
R( s )
-
Gp ( s )
Y( s )
E( s ) R( s )
Y( s )
Y( s ) K .Gp ( s )
.R ( s )
E( s ) R( s ) . 1
K .Gp ( s ) .H( s ) 1) .R ( s )
substitute Y(s)
1 E( s )
common denominator
SSE
lim s . s 0
K .Gp ( s ) ( H( s ) 1
and
1)
K .Gp ( s ) H( s ) R( s )
.R ( s )
If
K , Gp ( s ) , H( s )
numerically
69
Design Example
+
R( s )
-
Gp ( s )
Y( s )
Gp ( s )
(s
1 1 ) .( s
5)
1) Find K such that if r(t) is a unit step, then y(t) has the following characteristics
a)
ts
4. 2 18
b) PO = 5% (Percent Overshoot) 2) Find y(t) for r(t) = step 3) Find the steady state error for r(t) = step Calculate Pole location parameters.
PO 5 %
1 2
} plot
ts
4 .
18
Direct Calculation
dcl ( s ) s
2 . . n .s
6 .s s 3
18 j3
and
j3
} quadratic formula
Imag Axis
1 0
}
-5 -4 -3 -2 -1 0 1 2
-1 -2 -3 -4 -6
Y( s ) R( s )
=
K .Gp ( s ) 1 K. KGp ( s ) (s 1 1 ) .( s 5)
cl ( s )
Real Axis
70
K .Gp ( s ) Y( s ) R ( s ) 1 K .Gp ( s ) K. 1 1 ) .( s
( s) 1
K .Gp ( s )
0 s K 13 3 j3
1 5) s
=
(s
3 1 s
j3
Use Calculator
13 1 . ( s 1 ) .( s 5 ) Y( s ) s 13 1 ( s 1 ) .( s 5 ) Y( s ) 13 . 1 18 .s A .s (s 3) 1. s 3
2
13 s. s
2
6 .s
18
B ( 3) 1 1 s 3) s 3)
2 2 2
} PFE Formulation
A .s
B .s
1 . 2 s 18 1 18 s 1 18 .s 1 18
1 18
1 3
} PFE Tool
Y( s ) 13 .
1 . 18 ( s 1 . 18 ( s 1. e 18
3 t.
6 ( 3) 3 ( 3)
2 2
Y( s ) 13 .
3 (s
3 .t
3)
( 3)
y( t ) 13 .
cos ( 3 .t )
1. e 18
sin ( 3 .t )
y( )
13 18
y( 0 )
s .Y( s ) 0
71
E( s ) R( s )
Y( s )
}
.R ( s )
Definition
Y( s )
K .Gp ( s ) 1 K .Gp ( s )
}
.R ( s )
E( s ) R( s )
K .Gp ( s ) 1 K .Gp ( s )
substitute Y(s)
E( s )
1 .R ( s ) . K Gp ( s )
Simplify
SSE
lim s .E ( s ) s 0
1 lim s . . s 0 s 1
1 (s 13 1 ) .( s 5)
FVT
R( s )
1 s
SSE
5 18 e( t)
}
lim t
answer
lim t
( r( t)
y( t ) ) = t
lim 1
r( t) 13 18
lim t 5 18
y( t )
Another method
Note:
lim t
y( t )
lim s .Y( s ) s 0
13 18
72
R( s )
-
Gc ( s )
1 s .( s 1)
Y( s )
Find Let
Gc ( s )
3 j3
Gc ( s ) K 1)
Imag Axis
cl ( s ) 1
K.
1 s .( s
-1
}
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1
3 j3
-2
-3 -3.5
Real A xis
If
Gc ( s ) K . K. b
(s (s (s (s
a) b)
b> a> 0
cl ( s ) 1
a) . 1 b ) s .( s 1 ) 1 a
Imag Axis
-3+j3 0 -3-j3
Centroid RD 2
0 2
3 j3
-b
-c
-a
Real Axis
Note: a closed loop pole at "-c" has been injected By moving "a" and "b" around we can change the centroid and hence bend the root locus to go through 3 j 3 Note:
3 j3
How do we find the values for a, b and K which ensure that are the closed-loop poles ?
3 j3
73
Y( s ) R( s )
K( s a ) s .( s b ) .( s 1 ) s a 1 K. ( s b ) .s .( s 1 )
( s) 1
K.
( s a) . s ( s b ) .( s
=0
1)
j3
Angle K .
s .( s
( s a) b ) .( s
1)
j .3
= -180
} }
Angle Conditions
Angle
(s (s
a) . 1 . b) s ( s 1)
3
=
j .3 180
= - 180
Let
a 3
Angle ( s
b)
1
s 3
j .3
Angle s .
(s (s
1) 3)
j .3
Note :
b A Angle B Angle (
tan
11.3 deg
=>
Angle ( A) j ) tan
1
Angle ( B )
Selection of "a" 1) "a" is not unique 2) If "a" is picked too big, "b" will not exist (i.e., b may become negative) 3) If "a" is picked too small, the system might not exhibit the right dominant behavior (i.e., closed loop pole at -c might dominate the behavior)
74
K.
(s
s 3 18 ) .( s ) .( s
1)
j .3 1
K.
15
j .3 j .3 . 3 j .3 . 2
j .3
K.
K 79 ( approximately)
answer
Gc ( s ) 79 .
75
R( s )
-
Gc ( s )
Gp ( s )
Y( s )
Find
Gc ( s ) 1 s .( s
3 j3
Gp ( s )
1)
Gc ( s ) K .
s s
a b
K .( s a ) Y( s ) s .( s b ) .( s 1 ) R( s ) ( s a) s .( s 1 K. ( s b ) .s .( s 1 ) cl ( s ) s .( s 1 ) .( s
K .( s a ) 1 ) .( s b ) K .( s
a)
closed loop TF
Note: There are three closed loop poles given by the characteristic equation
b)
K .( s
a)
j
Imag Axis
-3+j3 0 -3-j3
-b
-c
-a
Real Axis
cl ( s ) s
(b
1 ) .s
(b
K) .s 3
a .K j .3 )
} }
c ) .( s c) . s (6
2
j .3 ) .( s 18 ( 18
6 .s
2
}
6 c) s 18 c
simplification
c) s
76
dcl ( s ) cl ( s ) 6 18 c b 1 K
6c b
aK 18 c
}
18 K K 78
a 3
K 6c 5 c b b 6c c 13
}
}
1 0 1
0 1 1
6 1 6
K . b c
0 5 18
b 18
Note we have discovered the third closed loop pole in the process (i.e., s = - 13)
c b
b 6c
18
18 . c a
eliminate K and b
c 6 .c
18
18 . c a
18 a
13
simplified expression
From root locus, "c" must be a positive number for the problem to make sense so if
c> 0
then
18 a
5> 0
therefore
a<
18 5
Any other selection of "a" forces "b" to be negative, and hence, the compensator would have an unstable pole.
77
Find the Structure of the Time Response For the Previous Problem Quickly
Y( s ) R( s ) ( s y( t ) A
78 .( s 13 ) . ( s B .e
13 t
3) 3) C .e
2 3 t.
R( s ) ( 3)
2
1 s
where
r( t)
is a step input
cos ( 3 t )
D .e
3 t.
sin ( 3 t )
A, B,C, D can be found with the PFE tool How can we find A ?
lim sY ( s ) s 0 lim t y( t )
78 .( 3 ) 1 ( 13 ) .( 18 )
so
lim t A 1
y( t )
A 1
78
Example
r( t)
+ -
y( t ) P( s )
H( s )
H( s )
s 1 ( s 4 ) .( s 2 ) Y( s ) P( s ) R ( s ) 1 P ( s ) .H ( s ) K
P( s )
s 3 s 0.1
cl ( s ) 1
K .P ( s ) .H( s ) 1
K.
s 3 . s 0.1 ( s
s 1 2 ) .( s
4)
j .a c
Imag Axis
0.5
-0.5
-1
j .a
-4 -2 0 2 4
-1.5 -6
Use coefficient matching to find the value of K which places closed loop poles at
Real Axis
j.a
3 ) .( s 4s 4 .K .s ( 0.8 3 3 .K 3 .K) 1)
cl ( s ) ( s cl ( s ) ( s cl ( s ) s cl ( s ) s
3 3
2 ) .( s 6s 8.6 s
4) 8 0.8 ( 8.6
K .( s K. s Ks
2 2
K) s
4 .K) .s
Simplification
79
cl ( s ) s
( 6.1 c) . s cs
2
K) s
2 2
( 8.6
4 .K) .s
( 0.8
3 .K)
actual denominator
dcl ( s ) ( s dcl ( s ) s
3
2 2
}
a c
a s
simplified
K c
2 4 .K a
3 .K a c
2
3 equations 3 unknowns
After multiplying equation 1) by equation 2), and equating the result to equation 3), we obtain one equation for K
( 6.1 4K 4K
2
4 K) 0.8 24.4 ) .K
3K a c 3K
52.46 0.8
51.66 0
} }
simplified
(K
1.94 ) .( K
or
6.657 ) 0 K 6.657
factored
K 1.94
K 1.94
For
Imag Axis
0.5
-0.5
-1
-1.5 -6
-4
-2
Real Axis
80
Gc ( s )
4 s .( s 2)
Y( s )
Design
Gc ( s )
s s
2 2
j .2 3 j .2 3
Gc ( s ) K .
s s (s (s
a b
Lead Compensator
Y( s ) R( s )
K. 1
a) . 4 b ) s .( s 2 ) ( s a) . 4 K. ( s b ) s .( s 2 )
cl ( s ) 1
K.
s s
a . 4 . b s ( s 2)
Imag Axis
j2 3
-b
-c
-a
j2 3
Real Axis
81
a 2.9 s 2.9 . s b (s
Angle
4 2 ) .( s )
180
j .2 . 3 4 .( s 2.9 ) s .( s 2 )
Angle ( s
b) s 2
j .2 3
180
Angle
j .2 3
tan
2 3 b 2
45
=>
b 5.46
Magnitude Condition
K.
s 2.9 . 4 s 5.46 s .( s 2 )
=>
K 4.68
j2 3
Y( s ) R( s )
(s (s
4 K (s b ) .( s 2 )
a) ( Ks
Ka ) 4
cl ( s ) s dcl ( s )
b ) .s
(2 b 2
4 K) .s
4 Ka
}
c)
2 (4
j .2 3 . s c) s
2
j .2 3 .( s 4 .c ) .s 16 .c
dcl ( s ) s
( 16
simplified
82
4 2 .b
c 2
b 4 .c
4 .K 16
16 .c 4 .K .a
Let
}
}
2 16 0
3 equations 4 unknowns
1 4 16 c b K
3 equations 3 unknowns
Imag Axis
j2 3
-b
-c
-a
j2 3
Real Axis
83
R( s )
K
-
(s
1 1 ) .( s
Y( s ) 5)
Let
R( s )
Y( s ) R( s )
1 s K.
1 ( s 1 ) .( s 5 ) 1 K. ( s 1 ) .( s 5 )
E( s ) R( s )
Y( s )
By Defintion
K. E( s ) R( s ) 1
1 ( s 1 ) .( s 5 ) R( s ) 1 K. ( s 1 ) .( s 5 )
E( s ) 1 K. (s
1 1 1 ) .( s 5)
.R ( s )
Simplify
SSE t
lim
e( t)
lim s .E ( s ) s 0 1
1 K 5
R( s )
1 s
84
Same Problem
Y( s ) R( s )
K. 1
1 ( s 1 ) .( s 5 ) 1 K. ( s 1 ) .( s 5 ) K 1 ) .( s
3 j3
cl ( s ) 1
4 3 2
(s
5)
Imag Axis
1 0
-1 -2 -3 -4 -6
-5
-4
-3
-2
-1
Real Axis
1 ( s + 1)( s + 5)
1 1
=1
s =3+3 j
=>
K 13
SSE
5 K 18 5
From the above problem, we can see that the SSE is fixed since we have selected a value of K to place the poles at
3 j3
85
Lag Compensation
( The preferred method for reducing SSE )
SSE 1 K
5 13 18 5
} For K = 13
Y( s )
R( s )
-
Gc ( s )
Gp ( s )
Let
R( s )
1 s 1 1 ) .( s c d 5) c> d> 0
Typical Lag Compensator Let
Gp ( s )
(s s s
K 13
Gc ( s ) E( s )
c 0.1
d 0.01
}
1
E( s ) R( s )
Y( s )
SSE ( s )
1 lim s . . s 0 s 1
1 K .Gp ( s ) .Gc ( s )
SSE ( s )
s 0.1 . s 0.01 ( s
1 1 ) .( s
}
5)
SSE ( s )
1 1 130 27 1 5
and
( approximately)
If
Gc ( s ) 1
with
K 13
SSE
1 ( approximately) 3
Gc ( s )
s 0.1 s 0.01
Question : Have we created a problem by adding a lag compensator ? That is, will the output still behave according to the dominant specifications ?
86
Lead/Lag Compensators
Example
R ( s)
GC ( s )
GP ( s )
Y ( s) Gp ( s ) = (s 1 1 ) .( s 5)
Design a lead/lag compensator that has 1) Dominent closed loop poles as s = - 4 +/- j5 2) SSE = 0.01 when R(s) = 1/s Design Lead Compensator to place "dominant" closed-loop poles at 4
j5
effects of the lag compensator. Then design lag to achieve SSE specification. Note - if we dont neglect the lag during the lead design, the problem becomes fairly difficult to solve. Let
G c( s)
=
K( s (s
a) b)
} Lead Compensator
}
closed loop TF
Y( s) R( s) ( s) 1
Gc ( s ) . G p ( s ) 1
G ( s ) .G ( s ) c p 1 1 ) .( s
K.
s s
a . b (s
5)
25 20 15
Imag Axis
10 5 0 -5
j5 -b -c -a - j5
Real Axis
87
Angle Condition
angle
s (s
1
10 1) ( s 5)
(s = -4 + 5j)
b) ( s 5
= -180
Let a = 10 Why ?
tan
19.3
b = 18.3
Magnitude Condition
K
s (s 18 ) ( s
10 1) ( s 5)
s=-4+5j
=1
K=
(s
18 ) ( s (s
1) ( s 10 )
5)
s=-4+5j
= 56.6
c d
Lag
SSE
lim s . s 0 1
G c( s) G p( s)
SSE 1 1 1 1 c d 6.3
1 56.6 ( 10 ) c ( 18 ) ( 1 ) ( 5 ) d 0.01
}
}
R( s)
1 s
SSE
c d
simplify
= 15.7
Gc ( s ) 56.6 .
s s
88
Why can we neglect the Lag dynamics when we design the Lead compensator ? That is, how close was the above calculation ? Let us re-examine the Lead calculation with the Lag Dynamics inserted
cl ( s ) =
56.6 .
s s
1 1 ) .( s
5)
= 0
j5
Angle Condition
s s 10 . s 0.1 . 18.3 .00625 ( s 1 1 ) .( s
should be equal to -180;
5)
j5
Magnitude Condition
56.6 s s 10 . s 0.1 . 18.3 s .00625 ( s 1 1 ) .( s 5)
should be equal to 1;
s 4 5j
however, it is equal to 0.973 The reason that the above calculations are still close is because which is close to
s s
0.1 .00625
= 0.9819
s = -4+5j
0.66 o
1 0
Hence, by selecting the lag compensator zero close to the j axis and the lag compensator pole relatively close to the lag compensator zero, we change the angle and magnitude calculation very little. The above approximation method is fairly accurate.
89
Example
R ( s)
GC ( s )
GP ( s )
Y ( s)
Design a lead/lag compensator that has 1) Dominant closed loop poles at s = -2 +/- 2j 2) SSE = 0.01 when R(s) =
Gp ( s ) 1 s
1 (s 1) ( s 3) ( s 10 )
Design Lead to place the "dominant" closed-loop poles at -2 +/- 2j while neglecting the effects of the Lag compensator.
G ( s) K.
c
s s
a b
Lead Compensator
Y( s) R( s) ( s) 1
G ( s ) .G ( s ) c p 1 K. s s a . b (s
G ( s ) .G ( s ) c p 1 ) .( s
}
1
closed loop TF
3 ) .( s
10 )
15
10
Imag Axis
-b -a
0
j2 - j2
-c
-5
-10
-15
-12
-10
-8
-6
-4
-2
Real Axis
90
Angle Condition
angle
s (s b) ( s
3.5 3 ) .( s 10 )
(s = -2 + 2j)
1) ( s
= -180
b = 4.38
Magnitude Condition
s (s 4.38 ) ( s 3.5 1) ( s 3 ) .( s 10 )
s=-2+2j
=1
=>
K 50
Lead
Lag
SSE
lim s .E ( s ) s 0 1 1 c d
lim s
1 0 1
=
G c( s) G p( s) 1 1 1.332 .c d
R ( s ) 0.01
SSE =
1 ( 50 ) ( 3.5 ) ( c ) 30 ( 4.38 ) d
= 74.32
= 0.01
"c" is typically chosen to be 0.1; hence, d = .001345 Why can we neglect the Lag dynamics when we design the Lead compensator ? That is, how close was the above calculation ? Let us re-examine the Lead calculation with the Lag Dynamics inserted
cl ( s )
50.0 .
s s
1 1 ) .( s 3 ) .( s 10 ) s 2 j2
= 0
91
Angle Condition
s
angle
1 1 ) .( s 3 ) .( s 10 ) s 2 j2
Magnitude Condition
50 . s s s 0.1 3.5 . . 4.38 s .001345 ( s 1 1 ) .( s 3 ) .( s 10 ) s 2 j2
should be equal to 1;
however, it is equal to 0.9034 The reason the above calculations are still close is because which is close to
s s
0.1
0.001345
=
s = -2+2j
0.9512 1.43707o
1 0
Hence, by selecting the lag compensator zero close to the j axis and the lag compensator pole relatively close to the lag compensator zero, we change the angle and magnitude calculation very little. The above approximation method is fairly accurate. Note: The transient response of the Lead / Lag compensator is similar to the Lead / No Lag compensator; however, it is clear that the steady state responses are very different
92
Lead/Lag Compensator
output response to a step input Note the similarities of the transient responses
93
94
b) d ) .( s e)
a ,b ,c ,d ,e
are
ab . H( s ) K . ecd
j 1 . b j 1 . d
1 j 1 . e 1
Bode Form
Magnitude Plot 1) Curve breaks { up / down } by 20 dB/dec at { a,b / c,d,e} Angle Plot 2) Curve breaks { up / down } by 45 degrees one decade before { a,b / c,d,e } and breaks { down / up } by 45 degrees one decade after { a,b / c,d,e } Initial Magnitude Plot Case 1: n=0 Calculate
H( j )
20 .log( H( j ) )
20 .n
intercept
dB / dec dB / dec
Kab cde
1 n
95
Initial and Final Phase Plot 1) Plot a pole / zero plot for the original transfer function 2) Calculate the angle from the picture at
Illustrative Example
Angle
s s
Initial Angle =
Final Angle =
a)
c) )
= 90 - 90 = 0 degrees
1 0.8 0.6
Imag Axis
0.4 0.2 0
Real Axis
96
s 1 H( s ) s 10
Magnitude Plot 1) Initial Magnitude => n = 0 =>
H( j )
1 . j 10 1 j 10 1
}
1 10
Bode Form
H( j ) 20 dB
Initial Magnitude =
20 .log
1 10
dB Conversion
2) Plot the negative of the poles and the zeros on the j axis zero at -1 3) pole at -10
Use an "up" arrow to denote a slope of +20 dB/dec and a "down" arrow to denote a slope of -20 dB/dec
H( j )
101 100 101 102
RD 0
Final Slope = 0 dB / dec
dB
-20 -40 Initial Magnitude Decade
4) Start at the initial magnitude and use arrows to draw the asymptotic plot
H( j ) dB
-20
101
100
101
102
+ 20 dB/dec -40
97
Phase Plot
for
s 1 H( s ) s 10
H( j )
1 . j 10 1 j 10 1
Bode Form
1 0.8 0.6
Imag Axis
0.4 0.2 0
10
}
0
-10
-8
-6
-4
-2
Real Axis
1) Initial Phase
Angle ( H( j ) )
0
Note :
Angle
j 1
Angle
j 10
= 0-0=0
see picture
Final Angle
Angle ( H( j ) )
Angle
j 1
Angle
j 10
Note :
= 90 - 90 = 0
see picture
98
2) Plot the negative of the poles and the zeros on the j axis zero at -1 pole at -10
3) Use an "up" arrow to denote a slope of +45 deg/dec and a "down" arrow to denote a slope of - 45 deg/dec
Angle ( H( j) ) deg
90 Decade
45
Initial Phase
Final Phase
10 1
4)
10 0
101
10 2
Frequency (rad/sec) Start at the initial phase and use arrows to draw the asymptotic plot
Angle ( H( j) ) deg
90 Initial Phase 45
45 deg/dec
D ecade
F inal P hase
- 45 deg/dec
10 1
10 0
10 1
10 2
F requency (rad/sec)
Use the Asymptotic Plots to draw the actual Bode Plot free hand (rough out the edges)
B o d e D i a g ra m s 0 -5 Phase (deg); Magnitude (dB) -1 0 -1 5 -2 0
50 40 30 20 10 10
-1
10
10
10
F re q u e n c y (ra d /s e c )
99
H( s )
10 ( s 10 ) s ( s 1 ) ( s 100 ) j 1 j 100 1
10 ( 10 ) . H( j ) 100
10 j j 1 1
}
}
Bode Form
Magnitude Plot 1) Initial Magnitude => n = 1 => Calculate the intercept 1 n 10 ( 10 ) 0 I 1 10 n 1 100
n 1
100
2) Plot the negative of the poles and the zeros on the j axis zero at -10 3) poles at 0 , -1 and -100
Use an "up" arrow to denote a slope of +20 dB/dec and a "down" arrow to denote a slope of -20 dB/dec
H( j )
20 0
dB
-20 -40 -60
}
102
I
100
101
Frequency (rad/sec)
4) Use the initial slope and the arrows to draw the asymptotic plot
H( j )
20
dB 0
-20 dB/dec -20 -40 -60 -20 dB/dec -40 dB/dec -40 dB/dec
100
101
10
101
Phase Plot
for
H (s) =
10 ( s 10 ) s ( s 1 ) ( s 100 ) j 1 j 100 1
H (s) =
10 ( 10 ) 100
10 j j 1 1
} }
Bode Form
1 0 .8 0 .6
Imag Axis
0 .4 0 .2 0
-0 .2 -0 .4 -0 .6 -0 .8 -1
100
10
R e a l A x is
1) Initial Phase
Angle ( H( j ) )
10 = 0 90 0 0
Angle
= 90
Final Angle
Angle ( H( j ) ) j 10
Angle
= 90 90 90 90 = 180
102
2) Plot the negative of the poles and the zeros on the j axis zero at -10 poles at 0 , -1 and -100
3) Use an "up" arrow to denote a slope of +45 deg/dec and a "down" arrow to denote a slope of - 45 deg/dec
Angle ( H( j ) )
10 1
90 0
10 0
10 1
10 2
10 3
deg
-90 -180
Initial Phase
-270
Decade
Final Phase
4)
Start at the initial phase and use arrows to draw the asymptotic plot
Angle ( H( j ) )
-45 -90 Initial Phase Final Phase - 45 deg/dec -180 -225 - 45 deg/dec
deg
-135
10 1
10 0
101
10 2
10 3
Frequency (rad/sec)
103
Use the Asymptotic Plots to draw the actual Bode Plot free hand (rough out the edges)
Bode Diagrams
-50
-100
10
10
10
10
10
10
Frequency (rad/sec)
1 s n
2
H (s) =
1 n
2
1 j n 1 j n 1
}
dB
Bode Form
H( j )
1 0
=
Initial Magnitude =
20 .log
1 n
2
dB Conversion
104
2) Plot the negative of the poles on the j axis 2 poles at - n 3) Use an "up" arrow to deonte a slope of +20 dB/dec and a "down" arrow to denote a slope of -20 dB/dec
H ( j )
20 0
dB
n /10
10n
100n
105
Phase Plot
for
H (s) =
1 s n 1 n
2 2
H (s) =
1 j n 1 j n 1
} }
Bode Form
1 0.8 0.6
Imag Axis
0.4 0.2 0
Real Axis
1) Initial Phase
Angle ( H( j ) )
Angle
j n
Angle
j n
1 0
= 0 0 = 0
Final Angle
Angle ( H( j ) )
Angle
j n
Angle
j n
= 90 90 = 180
106
2) Plot the negative of the poles and the zeros on the j axis 2 poles at -n 3) Use an "up" arrow to denote a slope of +45 deg/dec and a "down" arrow to denote a slope of - 45 deg/dec
Angle ( H( j ) )
n 10
90
10 n
100 n
deg 0
-90 -180 -270
Final Phase Initial Phase
4)
Start at the initial phase and use arrows to draw the asymptotic plot
n
90 0 -90 -180 -270
10
10n
100n
107
Use the Asymptotic Plots to draw the actual Bode Plot free hand (rough out the edges) In the following plot
n 1
Bode Diagrams
10
10
Frequency (rad/sec)
108
Example
Using the Bode Plot for determining Stability of a Closed-Loop System
R( s )
B
Y( s ) K H( s )
Y( s ) K .H( s ) R ( s ) 1 K .H ( s ) ( s) 1
100 80 60
H( s )
1 s 1 s 1 . 10 1 . s 100 1
} }
K .H ( s )
Imag Axis
j d
c 100
10
}
j d
0 20
-100
-120
-100
-80
-60
-40
-20
Real Axis
For
0 < K< K
c , j d , j d
Angle ( H( s ) )
= - 180
s j d
Magnitude Condition
}
= 1
be used to find
and
K . H( s )
s j d
109
and
d
=>
Angle ( H( s ) )
= - 180
s j d
Angle H jd
180
u.
H( s )
s j d
= 1
=>
u K . H j d
H( j )
and
Angle ( H( j ) )
for all
; hence
d
With
d
u
H j d
u
is
calculated as follows
1 H j d 1
H j d 20
} }
is in dB
is "regular" units
10
Bottom Line: The Bode Plot gives us a graphical solution for finding u
and
( s) 1
K.
1 s 1 s 1 . 10 1 . s 100 1
110
and
Magnitude (dB)
- 40 dB
}
-180 deg
-1 0 1 2 3
Phase (deg)
d = 20
10
10
Frequency (rad/sec)
H j d
40 dB
H j d 20
} }
}
10
in dB
For u
answer
Imag Axis
40 20 0 -2 0 -4 0 -6 0 -8 0
j d
c 100
10
j d
-100
-1 20
-10 0
-8 0
-60
-40
-20
20
R eal A xis
111
Y( s ) K H( s )
cl ( s ) 1
K .H ( s )
Magnitude (dB)
KH ( j )
Phase (deg)
KH ( j p )
-180
10
-1
10
10
10
10
Frequency (rad/sec)
K .H( j)
0 dB 180
Phase Margin -
pm 180
(If
Angle H j .p
unstable )
Gain Margin -
GM
112
The Bode design parameters are related to the root locus design parameters according to 1 2
p n . 4 .
2 .
1 2
}
pm < 60
pm tan
4 . 4 . 2 .
2 .
The phase margin is also approximately related to the damping ratio as follows
0.01 . pm
Remember
degrees for
degrees
Rule of thumb
Im{s}
x n
cos ( )
Re{s}
Hence,
pm
113
R( s )
B
Gp ( s )
Y( s )
Gp ( s )
1 s. s 2 1
Find K so that
pm 45
Outline of the Solution 1) Find the closed-loop denominator in the form 2) Plot
( s) 1
K .Gp ( s )
K .Gp ( j) pm
and
Assume K = 1
3) From i.e.
From given
pm
4)
and the
Angle Gp ( j)
plot , find
p p
5) Raise or lower magnitude plot so that it crosses the 0 dB level at the desired value of 6) Find the the desired value of
between the desired magnitude plot and the original magnitude plot at
p
desired + original = shift dB
7) Calculate K shift 20
K 10
114
Step 1
( s) 1
K. s.
1 s 2 1
Step 2
40 Magnitude (dB) 20 0 -20
K 1
Phase (deg)
}
10
-1
KGp ( s )
10
10
Frequency (rad/sec)
Step 3
pm 45
115
Angle Gp jp
135
Find
40 Magnitude (dB) 20 0 -2 0
For K = 1
Phase (deg)
-1 0 0 -1 2 0 -1 4 0 -1 6 0 10
-2
-1 1 8
-1 3 5
}
10
-1
For
pm
K 1 180
118
pm 62
10
10
Step 5
F r e q u e n c y (ra d /s e c )
O r ig in a l M a g n itu d e P lo t 40 20 -9 d B
D e s ir e d M a g n itu d e P lo t
}
-1 3 5
-2 0
Magnitude (dB)
Phase (deg)
-1 0 0 -1 2 0 -1 4 0 -1 6 0 10
-2
10
-1
10 F r e q u e n c y (r a d /s e c )
10
Step 6
The difference between the desired magnitude plot and the original magnitude plot is
9 dB
; therefore
Step 7
K 10
=>
K 10
=>
K 2.8
116
Y( s ) K Gp ( s )
Gp ( s )
1 s. s 2 1
pm 45 deg
K .Gp ( s ) Y( s ) R ( s ) 1 K .Gp ( s ) K. 1 2 s .( s
Y( s ) R( s )
2) 2 K. s .( s 2 ) 2 s .( s 2)
}
}
( s) 1
2 1.5 1
K.
jb
Imag Axis
0.5 0
-0.5 -1
}
-jb
0 1 2
-1.5 -2 -3
-2
-1
Real Axis
If Note:
pm 45 cos ( )
=>
0.45
117
0.45
cos ( )
=>
63 deg
Formulas =>
b 1 sin ( 63 ) sin ( 27 )
Desired closed loop poles
}
2 1
Law of sines
b 2
1 j2
Magnitude Condition
K.
2 s( s 2) s 1 j2
= 1
=>
K 2.5
Why is this number different from K = 2.8 obtained from the Bode Design ?
pm 45 deg
0.45
desired closed loop poles =
1 j2
118
Example
R( s )
B
Y( s ) K Gp ( s )
Find K so that
pm 45 deg 1 s 1 . 10 K. 1 . s 100
Gp ( s )
s 1
Step 1
( s) 1
1 s 1 s 1 . 10 s 1 . 100 1
Step 2
0 Magnitude (dB)
-50
For
K 1
-100
10
10
10
10
Frequency (rad/sec)
Step 3
pm 45
119
Angle Gp jp
135
Find
0 Magnitude (dB)
-50
For
K 1
-100
-135 desired
10
0
p
10
1
}
10
2
For
pm
K 1 180
pm 180
10
Frequency (rad/sec)
Step 5
Original M agnitude Plot -20 dB Magnitude (dB) Desired M agnitude Plot
{0
-50 -100
-135
p
10 10 10
2
10
Frequency (rad/sec)
Step 6
The difference between the desired magnitude plot and the original magnitude plot is
desired
Step 7
original 0
shift 20
20
20 dB
20 20
; therefore
shift
20 dB
K 10
=>
K 10
=>
K 10
120
121
Y( s ) K Gp ( s )
Y( s ) R( s ) ( s)
KGp ( s ) 1
=
KGp ( s ) 1 KGp ( s )
Closed loop TF
Im K .Gp ( j )
versus
Re K .Gp ( j )
using the Nyquist contour. (Assume K = 1) 2) Apply the Nyquist Criterion to the Nyquist Plot and then determine the stability of the closed-loop system.
Nyquist Criterion P0
denotes the number of poles of
Gp ( s )
1 180
If the contour in the s-plane is CW, then CW is the positive direction. i.e., 1 CW
N =1 N =1
1 CCW
N =1 N =1
If the contour in the s-plane is CCW, then CCW is the positive direction. i.e., 1 CCW 1 CW
Zc
Zc N
P0
122
Example
Draw the Nyquist Plot for for the following s-plane contour. Step 1) Let
Im {s}
H( s )
1 s 1 1
Nyquist Contour
s j
H( j . )
j .
1
R
R Re{s}
Step 2) Plot a pole/zero plot for H(s) on the s-plane contour picture and include measurement scheme to the j - axis along with the necessary critical points.
Step 3) Enter a magnitude/phase table entry for each critical point. Point r mag Phase
Im {s}
A B C H ( j )
1 0 0 1 j
0 90 0 r 1 1 1)
-1 A C Re{s}
Angle( H ( j ) )
Angle( j
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half.
Im{H(j )}
B,C
A r Re{H(j )}
123
Example
Draw the Nyquist Plot for for the following s-plane contour. Step 1) Let
H( s ) 1 j ( j .
1 s( s 1)
Im{s}
Nyquist Contour
s j H( j . )
1)
R
0
R Re{s}
Step 2) Plot a pole/zero plot for H(s) on the s-plane contour picture and include measurement scheme to the along with the necessary critical points. Step 3) Enter a magnitude/phase table entry for each critical point. Point r mag Phase
axis
Im{s}
A B C D H ( j )
0 0
0 90 180 0 1 j . j 1 Angle( j 1)
Im{H(j )}
B
"Blow-up"
Im {s} B 90
o
A -1
D Re{s}
A Re{s}
Angle( H ( j ) )
Angle( j )
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half.
-1
C,D
A r Re{H(j )}
r=
124
Y( s ) K Gp ( s ) Gp ( s ) s .( s
1 1 ) .( s
10 )
Im{s}
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects closed-loop stability.
Nyquist Contour
K .Gp ( s ) Y( s ) R ( s ) 1 K .Gp ( s ) K.
( s) 1
s .( s
1 1 ) .( s
10 )
Nyquist Form
R 0
Re{s}
Step 1) Let
s j
K .Gp ( j ) Gp ( s )
K ( j ) .( j j 1 ) .( j
axis
10 )
Assume K = 1
picture and include measurement scheme to the along with the necessary critical points. Step 3) Enter a magnitude/phase table entry for each critical point.
Im{s}
r
point mag
phase
A B C D
0 0
0 90 270 0 1 j . j 1 . j 10 Angle( j )
"Blow-up"
Im {s} B 90
o
B A -10 -1
D Re{s}
A Re{s}
Gp ( j )
Angle Gp ( j )
Angle( j
1)
Angle( j
10 )
125
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half. point
r
mag
phase
A B
Im{Gp(j )}
0 0
0 90 270 0
C D
C,D
A r Re{Gp(j )}
r=
Im{KGp(j )}
-1
C,D
A r Re{KGp(j)}
r=
Larger K B
As K is increased, the shape of the Nyquist plot does not change because K is a real number (i.e., K only affects the magnitude (size); K does NOT affect the phase (shape))
126
Im{KGp(j )}
Im {s}
-1
C,D
A r Re{KGp(j )}
-10
-1
Re{s}
r=
Nyquist Criterion
1) Find
P0
=>
P0 0
( Number of poles of
Gp ( s )
2) Determine the sign notation for the encirclements. Contour is CW so, CW encirclements are positive and CCW encirclements are negative
3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 0 2
P0
Stability Yes No
small K big K
0 2
4) Understanding the above table: For small K, no closed loop poles are inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is stable For big K, two closed loop poles are inside the s-plane contour; hence, for big K, the closed loop system is unstable
127
15
10
Gp ( s )
s .( s
1 1 ) .( s
10 )
Imag Axis
}
-12 -10 -8 -6 -4 -2 0 2 4
( s) 1
K .Gp ( s )
-5
-10
small K: closed loop system is stable big K: closed loop system is unstable
-15
Real Axis
N
small K big K
Zc 0 2
Stability Yes No
0 2
128
d H(j ) -1 pm
Im{H(j )} Re{H(j )}
unit circle
-1
Nyquist Plot
pm phase margin H j . p 1
1 H j .
1 d
Nyquist Plot contains the same type of performance specifications as the Bode Plot and the Root Locus
129
Im{s}
Y( s ) K H( s )
Nyquist Contour
Re{s}
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects closed-loop stability.
H( s )
s s
1 1
We can see from the root locus that the system goes unstable
Y( s ) K .H ( s ) R ( s ) 1 K .H ( s ) ( s) 1
Step 1) Let
K.
s s
1 1
Nyquist Form
Imag Axis
s j
K .( j 1 ) K .H ( s ) ( j 1 )
-0.2
Assume K=1
H ( s)
Real Axis
axis
Im {s}
r
point mag
phase
Nyquist Contour
A B C
1 1 1
180
A C 1 Re{s}
0 0
-1
H ( j )
j j
1 1
Angle( H ( j ) ) Angle( j
1)
Angle( j
1)
130
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half.
r
point mag
phase
Im{H(j )}
A B K 1
1 1 1
180 0 0
A -1
B,C r Re{H(j )}
Im{s}
Nyquist Contour
A -1 1
CW + CCW -
C Re{s}
As K is decreased below 1, the Nyquist Plot intersects the Re{j } axis before -1 As K is increased beyond 1, the Nyquist Plot intersects the Re{j } axis beyond -1
Nyquist Criterion
1) Find
P0
=>
P0 0
( Number of poles of
H ( s)
2) Determine the sign notation for the encirclements. Contour is CW so, CW encirclements are positive and CCW encirclements are negative 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 0 1
P0
Stability Yes No
small K big K
0 1
4) Understanding the above table: For small K, no closed loop pole is inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is stable For big K, one closed loop pole is inside the s-plane contour; hence, for big K, the closed loop system is unstable
131
Y( s ) K H( s )
Im {s}
Nyquist Contour
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects H( s ) closed-loop stability. Step 1) Let
Re{s}
s s
1 1
s j K .( j ( j 1) 1)
K .H ( s )
} K=1
H ( s)
Assume
axis
r
point mag
phase
Im {s}
Nyquist Contour
A B C
1 1 1
180 0 0
C -1 A 1
Re{s}
H ( j )
j j
1 1 1) Angle( j 1)
Angle( H ( j ) ) Angle( j
132
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half. point
r
mag
phase
Im{H(j )}
A B K 1 C
1 1 1
180 0 0
A -1
B,C r Re{H(j )}
CW CCW +
Im {s}
Nyquist Contour
As K is decreased below 1, the Nyquist Plot intersects the Re{j } axis before -1 As K is increased beyond 1, the Nyquist Plot intersects the Re{j } axis beyond -1
-1
Re{s}
Nyquist Criterion
1) Find
P0
=>
P0 1
( Number of poles of
H ( s)
2) Determine the sign notation for the encirclements. Contour is CCW so, CW encirclements are negative and CCW encirclements are positive 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 1 1 0
P0
Stability Yes No
small K big K
4) Understanding the above table: For small K, one closed loop pole is inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is stable For big K, no closed loop pole is inside the s-plane contour; hence, for big K, the closed loop system is unstable
133
Im{s}
Y( s ) K H( s )
Nyquist Contour
R 0
Re{s}
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects closed-loop stability.
H( s )
s .( s
1 1 ) .( s
2)
Y( s ) K .H ( s ) R ( s ) 1 K .H ( s ) ( s) 1 K. 1 s .( s
1 ) .( s
2) ( j ) .( j
}
K
Step 1) Let
s j
K .H ( s )
1 ) .( j j
2)
} K=1
Assume
H ( s)
picture and include measurement scheme to the Step 3) Enter a magnitude/phase table entry for each critical point.
Im{s}
r
point mag
Nyquist Contour
phase
A B C D
0 0
0 90 270 0
"Blow-up"
Im {s} B 90
o
B A -2 -1
D Re{s}
A Re{s}
H ( j )
Angle( H ( j ) )
134
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half. point
r
mag
phase
Im{Gp(j )}
K 1
A B C
0 0
0 90 270 0
-1
C,D
A r Re{Gp(j )}
r=
Im{s}
CW + CCW -
Nyquist Contour
B
-2 -1
B A
D Re{s}
Nyquist Criterion
1) Find
P0
=>
P0 0
( Number of poles of
H ( s)
2) Determine the sign notation for the encirclements. Contour is CW so, CW encirclements are positive and CCW encirclements are negative 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 0 2
P0
Stability Yes No
small K big K
0 2
4) Understanding the above table: For small K, no closed loop pole is inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is stable For big K, two closed loop poles are inside the s-plane contour; hence, for big K, the closed loop system is unstable
135
Im {s}
Y( s ) K H( s )
R
Nyquist Contour
Re{s}
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects closed-loop stability. H( s ) Step 1) Let
s j K ( j ) .( j 1 ) .( j 2)
s .( s
1 1 ) .( s
R 0
2)
K .H ( s )
} K=1
Assume
H ( s)
axis
Im {s}
Step 3) Enter a magnitude/phase table entry for each critical point.
r
point mag
phase
Nyquist Contour
B D -2 -1 A Re{s}
A B C D
0 0
0 90 270 540 1
"Blow-up"
Im {s} B 90
o
A Re{s}
H ( j )
j . j
1 . j
2 1) Angle( j 2)
Angle( H ( j ) )
Angle( j ) Angle( j
136
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half. point
r
mag
phase
Im{Gp(j )}
K 1
A B C
0 0
0 90 270 540
-1
C,D
A r Re{Gp(j )}
CW CCW +
D
Im {s}
r=
Nyquist Contour
-2 -1
Re{s}
Nyquist Criterion
1) Find
P0
=>
P0 3
( Number of poles of
H ( s)
2) Determine the sign notation for the encirclements. Contour is CCW so, CW encirclements are negative and CCW encirclements are positive 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 3 2 1
P0
Stability Yes No
small K big K
4) Understanding the above table: For small K, all three closed loop poles are inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is stable For big K, only one closed loop pole is inside the s-plane contour; hence, for big K, the closed loop system is unstable
137
R( s )
B
Y( s ) K H ( s) (s
2 s .( s
Im{s}
H( s )
Im{H(j)}
2 ) .( s 1 ) .( s
4) 3 ) .( s 5)
K 1
-b -1 -1/a
Nyquist Contour
Re{H(j)}
R 0
Re{s}
r=
5
Stable region
4 3
Imag Axis
2 1 0
This is how you might be tempted to draw the root locus without the Nyquist plot Does the Nyquist Plot concur ? 1) Find
-1 -2 -3 -4 -5 -7 -6 -5 -4 -3 -2 -1 0 1 2
P0
P0 0
Real Axis
2) Determine the sign notation for the encirclements. Contour is CW so, CW encirclements are positive and CCW encirclements are negative 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc Zc 2 0 2
Zc N
Stability No Yes No
P0
N
small K medium K big K
2 0 2
=>
138
N
small K medium K big K
Zc 2 0 2
Stability No Yes No
2 0 2
4) Understanding the above table: For small K, two closed loop poles are inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is unstable For medium K, no closed loop poles are inside the s-plane contour. From the contour, we can see that this means that for medium K the closed loop system is stable For big K, two closed loop poles are inside the s-plane contour; hence, for big K, the closed loop system is unstable The above table assists in the Root Locus sketch
5 4 3
* K2
}
} small K
large K
Imag Axis
2 1 0 medium K
{
K
* 1
-1 -2 -3 -4 -5 -7 -6 -5 -4 -3 -2 -1
Real Axis
From the above root locus, we can conclude that the system is stable for
* K1* < K < K 2
K1* =
1 b
and
* K2 =a
-b -1
-1 /a
139
R( s )
B
Y( s ) K Gc ( s ) Gp ( s )
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects closed-loop stability.
Im{s}
Nyquist Contour
Gp ( s )
Assume
1 s .( s 1) Gc ( s ) 1 K. 1 s .( s 1)
}
and
K 1
( s) 1
} for the
R 0
Re{s}
Step 1) Let
s j
K .Gp ( j ) Gp ( s )
K j .( j 1 ) j
axis
Assume K = 1
picture and include measurement scheme to the Step 3) Enter a magnitude/phase table entry for each critical point.
Im{s}
r
point mag
Nyquist Contour
A B C D
0 0
180 270
B A 1
D Re{s}
90
A R e{s}
180 0 1 j . j 1
Gp ( j )
Angle Gp ( j )
Angle ( j )
Angle ( j 1 )
140
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half.
r
point mag
phase
Im {G p (j )}
0 0
B
B C
C,D A -1 r
r=
Re{G p (j )}
D
Nyquist Plot with K=1
Im{s} C
CW + CCW -
B A 1 D Re{s}
Nyquist Criterion
1) Find
P0
=>
P0 1
( Number of poles of
Gp ( s )
2) Determine the sign notation for the encirclements. Contour is CW so, CW encirclements are positive and CCW encirclements are negative 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 2 2
P0
Stability No No
small K big K
1 1
4) Understanding the above table: For small and big K, two closed loop poles are inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is always unstable
141
R( s )
B
Y( s ) K Gc ( s ) Gp ( s )
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects closed-loop stability. Let
Gc ( s ) 2 s K. 2s s .( s
1 1 1)
( s) 1
Step 1) Let
s j
( 2 j 1 ) j .( j 1 )
}
C
Assume K = 1
picture and include measurement scheme to the Step 3) Enter a magnitude/phase table entry for each critical point.
Im{s}
r
point mag
phase
B A D 1 Re{s} -1/2
"Blow up"
Im {s} B 90
o
A B C D
0 0
180 270 90 0
A R e{s}
Gp ( j ) .Gc ( j )
Angle Gp ( j ) .Gc ( j )
142
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half.
r
point mag
phase
A B C
0 0
180 270 90 0
Im {Gp(j)Gc(j)}
-a C,D A -1 r Re{Gp(j)Gc(j)}
r=
CW + CCW -
B A -1/2 1
D Re{s}
Nyquist Criterion
1) Find
P0
=>
P0 1
( Number of poles of
Gp ( s ) .Gc ( s )
2) Determine the sign notation for the encirclements. Contour is CW so, CW encirclements are positive and CCW encirclements are negative 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 2
P0
Stability No Yes
small K big K
1 1
4) Understanding the above table: For small K, both closed loop poles are inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is unstable For big K,no closed loop poles are inside the s-plane contour; hence, for big K, the closed loop system is stable Stable for
K>
1 a
143
Example
R ( s )
B
Im{s}
Y( s ) K H( s )
Nyquist Contour
Re{s}
Use Nyquist Plot, Contour and Criterion to discuss quantitatively how the control gain K affects closed-loop stability.
H( s )
(s
s 1 1 ) .( s 10 )
Step 1) Let
s j
K .H( s )
K .( j 1 ) ( j 1 ) .( j 10 ) j
} Assume K=1
H ( s)
picture and include measurement scheme to the Step 3) Enter a magnitude/phase table entry for each critical point.
Im{s}
r
point mag
phase
A B C H( j )
0.1 0 0
1 . j
Angle ( H( j ) ) Angle ( j
144
Step 4) Use critical points to draw Nyquist Plot. Step 5) Nyquist plot is symmetric about real axis. Draw other half.
r
point mag
phase
Im{Gp(j)}
A K 1 B C
B,C
Im{s}
0.1 0 0
180 90 0
CW + CCW -
-1
A -0.1
Re{Gp(j)}
A -10 -1 1
C Re{s}
Nyquist Criterion
1) Find
P0
=>
P0 0
( Number of poles of
H ( s)
2) Determine the sign notation for the encirclements. Contour is CW so, CW encirclements are positive and CCW encirclements are negative 3) Find N (i.e., the number of encirclements of (-1,0) for different values of K Also compute
Zc N
Zc N Zc 0 1
P0
Stability Yes No
small K big K
0 1
4) Understanding the above table: For small K, no closed loop pole is inside the s-plane contour. From the contour, we can see that this means that for small K the closed loop system is stable For big K, one closed loop pole is inside the s-plane contour; hence, for big K, the closed loop system is unstable For
0 < K< 10
145
146
an y
(n)
an
1
bn
. ( n 1) 1 y . u( n 1 ) b
.......................
n 2 . u(n 2)
. a1 y
( 1)
a0 y =
............... b0 u
Typical DE
This type of system (i.e., a linear, causal, time invariant system, can also be written as follows
d x A. x dt y C. x
B.u
A is an n x n matrix C is 1 x n vector
B is n x 1 vector
where y(t) denotes the system output u(t) denotes the system input x(t) - an n x 1 state vector is used to account for "derivatives" in the system.
The "n" dimensions of the matrix equation takes care of the "n" derivatives found in the original "nth" order scalar differential equation.
147
(3)
a1 . y
( 2)
a2 . y
(1)
a3 . y b1 . u
( 2)
b2 . u
(1)
b3 . u
n 3
third-order DE
d x dt 1 d x2 dt d x dt 3
or
0 0 a3
1 0 a2
0 1 a1
.
x1 x2 x3
0 0 1
.u
x1 y b3 b2 b1
.
x2 x3
& = Ax + Bu x y = Cx
}
b2 s
2
From the original differential equation, we can obtain the transfer function as
Y( s ) U( s )
b1 s s
3
b3 a3
a1 s
a2 s
}
a3 . x1
We now calculate the transfer function from the state equation by decomposing the state equation into its scalar components
d x dt 1
x2
d x dt 2
x3
d x = dt 3
a2 . x2
a1 . x3
y b3 . x1
b2 . x2
b1 . x3
&1 = x2 x &2 = x3 x
Therefore
Direct Substitution
148
Y( s ) X1 ( s ) x1
( 3)
b1 s
b2 . s 1
b3
a3 . x1
a2 . x1
(1)
a1 . x1
(2)
u}
Direct Substitution
X1 ( s ) U( s ) s
3
1 a1 . s
2
a2 . s
} a3
b1 s b2 . s b3 Y( s ) . X1 ( s ) Y ( s ) } X1 ( s ) U( s ) U ( s ) s 3 a . s 2 a . s a 1 2 3
Cross Multiply
Hence, the state space form has the same transfer function as the original differential equation. There are an infinite number of state space forms for a given differential equation. The state space form for this example is called Reachable Canonical Form (RCF).
149
Example
Put (4)
(4)
6y
(3)
(2)
= (2)
3u
(3)
7u
(1)
( 1)
into RCF.
( 0) y
(3)
6y
( 2)
( 0) y
(1)
( 0) y
} }
Output side of DE
3.u
( 0) u
( 7) u
( 0) u
Input side of DE
& = Ax + Bu x y = Cx
0 1 0 0 0 0 1 0 -6 0 0 1 0
x -4x1 A-4x4
B-4x1 C-1x4
n 4
A=
0 0 0
B=
0 0 0 1
C=
Step 1
First place zeros everywhere except for the circled areas. The circled areas are the (i) super diagonal and the last row of "A", (ii) the last element of "B", and (iii) the entire "C" vector Step 2 Put "ones" in the Super Diagonal. Step 3 Put a one in the last enty of B. Step 4 Take the last "n" coefficients of the output side of the DE, reverse their order, make them negative, and then put them in the last row of A. Step 5 Take the "n" coefficients of the input side of the DE, reverse their order, and then put them in C.
150
Z( s ) L R2
+ u( t ) _
x +
y( t )
_ x
R1
1 Cs
Inductor Impedance =
Ls
Y( s ) Z( s ) U ( s ) Z ( s ) Ls Z( s )
=
}
1 Cs
Voltage Division
R1
||
R2 R1 Cs 1 Cs
R1 R2 Z( s ) R1 Z( s ) R2
}
R1 1
CR1 R2 s C R1 R2 s
Simplify
Y( s ) U( s ) s 2 R 1 R1
=
sR1 R2 C R2 LC
=
R1 R1 R2 C
=
}
R1
,
s L R2
3.8 H ,
C = .066 F
Typical Values
Y( s ) .2 . s 3 H( s ) 2 U( s ) s 4.s 3
1 0 &= x x + 3 4
y =[ 3 0.2] x
0 1u
x x = 1 x2
151
Matrix Review
Transpose of a matrix Exchange rows and columns
(A
Symmetric Matrix
B) T A
T
( AB )
B A
Standard Operations
} By Definition
Determinants
2x2
a1 a2 b1 b2 a1 a2 a3
=
a1 .b2
b1 .a2
3x3
b1 b2 b3= c1 c2 c3
a1 b2 c3 c1 b2 a3
b1 c2 a3 b1 a2 c3
c1 a2 b3 a1 b3 c2
Inverse of a Matrix
A .A A
1
Note
exists if
adj( A) A
1
( AB )
2x2
1 1
A d c
A b a hf gf) ge . 1 A
1 T
Standard Operations
a b c d a b c
ei A
1
( bi ai ( ah
hc ) gc gb )
bf ( af ae
ec 1 dc ) . A db
3x3
d e f g h i
( di dh
Rank of a Matrix A matrix "A" is called rank "m" if the number of linearly independent columns is equal to "m" x, y and z are linearly independent vectors if
x y
and
y z
152
Eigenvalues of a Matrix
I
n
A
1
0 ...... an
a1 .
(A
I) .v 0
A I
0 6 A
2
1 5 0 0 5 2
0 6 6 ( 2 0 6 1 5
1 5 3 ) .(
6 2) 0
1
=
}
. x y 1 3 . x y 1 2 0
Find roots
1 A
1 I .v1
3 0
0 3
.v
3
1
1 2
v 1
2 I .v2 0 6 1 5 2 0 0 2 .v 2
2
Eigenvector for
1 3
v 2
Note any multiple of
Eigenvector for
v1
or
v2
would be an eigenvector
2 6
or
2 4
} Examples
153
m n
(A 1 k p
.I)
m }
Nk
= Rank
(A
I )
k 1 Rank
(A
I ) }
Nks
xp
(A
p I) .xp 0
(A
I )
.x
then find
xj ( A
I) .xj
1 where
(j p
1,p
2 , ....... , 1 )
Nk
by 1
Nk 0
If not continue to step 5 Step 5: Find the largest k for which Find a vector
Nk
xj ( A
I) .xj
1,
2 , ...... , 1
154
Example
1 1 2 A 0 1 3 0 0 1
Repeated Roots Step 1:
I A 0 0 n 3
1 1 0 m 3 (A
2 3 1 1 I )
p
1)
Gives Eigenvalues
}
m 3
2
Eigenvalue
n I )
3 0
0 1 2 p 1 (A I )
1
0 0 3 Rank 2 p 2 (A 0 0 0 0 0 0 Rank 0
k 0 1 2 1
0 0 3 0 0 0 0 0 0
Rank 1
p 3 (A
I )
0 0 0 0 0 0 I ) I ) I ) I )
p 3 I )
1 2 3 k
} answer
Step 2:
Rank .( A I ) I ) I )
for
1 k p 3
3 2 1
2 1 1 1 0 1
but
Step 3:
Find a
xp
such that
p I) .xp 0
(A
I )
1.
xp 0 x . y z 0 0 0
0 0 0 (A I) .x3
3
x z
0 0 3
but
0 0 0 . y 0 0 0
(A
I) .x3
2
0 0 0 0 0 0
j 1 0
so a possible solution is
x3
0 1
First Eigenvector
155
Now find
xj ( A I) x3
I) .xj
for
j p I) x2
1,p
2 , ..... , 1
for j = 2, 1
p 3
x2 ( A
x1 ( A
j 2 0 x3 0 1 0 1 2 x2 0 0 0 0 1 2 3 0 Nk
j 3
0 1 2 x1 0 0 0
2 0
3 0 0
0 0 3 . 0
0 0 3 . 3
by one, all the results become zero. Hence we stop at this stage
x3 , x2 , x1
156
Example
1 1 2 A 0 1 3 0 0 2 1,m 2
Non Repeated Root
I A 0 0
1 1 0
2 3 2 (
2 1 ) .(
2) 0
Gives Eigenvalues
2,m 1 1 0 0 1 0
}
2 0
Eigenvalues
(A
I) .v 0
x z
0 0 0 1 I )
p
1 3 . y
3 1
First Eigenvector
Repeated Root
n 3 and m 2
Step 1:
Rank ( A
m 3
2 1
0 1 2 p 1 (A I )
1
0 0 3 0 0 1 0 0 5
Rank 2
(A
I )
0 0 3 0 0 1
Rank 1
p 2
} answer
157
Step 2:
I ) I ) I )
k 0 1
Rank .( A I ) I )
I )
1 2
for 1 k p 2 2 1 1 1
but
Rank .( A Rank .( A (A
3 2
Step 3:
Find a
xp
such that
p I) .xp 0
(A
I )
1.
xp 0
1,p 2 0 0 5 (A I) x2
2.
x z
0 0 0 0
but
0 1 2 (A I) x2
1.
x z
0 0 0
0 0 3 . y 0 0 1
0 0 3 . y 0 0 1
so a possible solution is
x2
for
1 0
First Eigenvector
Now find
xj ( A
I) .xj
j p
1,p
2 , .... , 1
p 2
x1 ( A
I) x2 0 0 Nk 1 0 0
j 1
0 1 2 x1 0 0 1
Step 4: After reducing all
0 0 3 . 1
Second Eigenvector
by one, all the results become zero. Hence we stop at this stage
x2 , x1
158
d x ax dt s .X( s ) X( s )
bu
x( 0) b .U( s ) s a
a .X( s ) b .U( s ) x( 0) s a
Simplify
b x (t ) = e at x(0) + '1 U ( s) s a
Let H ( s ) =
b h(t ) = e at b sa
} } }
Convolution Property
Convolution Integral
t x( t ) e
at
x( 0) 0
exp ( a ( t
) ) .b .u ( ) d
Input Contribution
d x Ax dt
Claim that
Bu
At
}
x( 0)
x - n x 1, t 0
A - n x n, B - n x 1
x( t ) e
exp ( A .( t
) ) .B .u ( ) d
exp ( A .t ) is a matrix
159
To Prove the above claim, we need three pieces of mathematical machinery 1) What is exp (A t) when "A" is a matrix ?
exp ( a .t ) 1 exp ( A .t ) I
2) Note
a .t A .t
2 2 a .t 2!
3 3 a .t 3!
......
} }
A .t 2!
2 2
A .t 3!
3 3
.....
d exp ( A .t ) 0 dt
.......
d exp ( A .t ) A I dt
Note:
A .t
A .t 2!
2 2
A .t 3!
3 3
....
A .exp ( A .t )
d exp ( At ) exp ( A .t ) .A dt
Why ?
because AB = BA if B = A
3)
a ( x) a2 ( x ) F ( x, u ) d 2 d d du + F ( x, a 2 ( x )) a2 ( x ) F ( x, a1 ( x )) a1 ( x ) F ( x, u )du = dx dx dx a1 ( x ) a1 ( x ) x
160
d x A .x dt x( t ) e
At .
B .u t x( 0) 0 exp ( A .( t ) ) .B .u ( ) d
} }
Matrix equation
Claimed Solution
t d } Insert the solution into LHS exp ( A .( t ) ) .B .u ( ) d dt 0 Take time t d At A .exp ( A .( t ) ) .B .u ( ) d Bu ( t ) } derivative x Ae x ( 0 ) and use dt Leibnitz Rule 0 t At solution . . A x B u Ae x ( 0 ) A .exp ( A .( t ) ) .B .u ( ) d Bu ( t ) } Insert into RHS 0 d d At . x e x( 0) dt dt
The right sides of both the above equations are equal ; hence, we have obtained a solution to
Note since
y C .x
& = Ax + Bu x
the solution for y(t) is given by
y( t ) C .e
At
x( 0)
C.
t 0
exp ( A .( t
) ) .B .u ( ) d
Output Solution
So if we have a system, given u(t), we can find y(t) with the above formula All we need to find is
161
exp ( A .t ) =
1 ( sI A)
By Definition
exp ( A .t )
exp ( A .t )
'{e at } = (s a )
}
1
If "a" is a scalar
'{e At } = (sI A)
exp ( A .t )
Example
Find
with
0 3
1 4 1 s 3 s
2
1 1 s 0 0 exp( At ) = ' 0 s = 3 4 1
1 4
}
3 (s
Definition
exp ( A .t ) =
1 s 3
4 1
. 1 s ( s)
where
( s) s
4s
1 ) .( s
3)
Take Inverse
exp ( A .t ) =
1 (s (s
s 4 1 ) .( s 3 . 1) ( s
3) ( s 3) ( s
1 1 ) .( s s 1 ) .( s
3) 3) 0.5 ( s 3) 1.5 ( s 3)
} }
1 exp ( A .t ) =
1.5 ( s 1) 1.5 ( s 1)
exp ( A .t )
3. exp ( t ) 2 3. exp ( t ) 2
1. exp ( t ) 2 1. exp ( t ) 2
1. exp ( 3 .t ) 2 3. exp ( 3 .t ) 2
162
d x dt
0 3
1 4
.x
0 1
.u x( 0) 0 0
y = [3 0.2] .x
y( t ) C .e
At
}
}
x( 0)
C.
t 0
exp ( A .( t
) ) .B .u ( ) d
Definition
y( t ) C .
t 0
exp ( A .( t
) ) .B .u ( ) d
x( 0)
0 0
1 t 1 3t 3 t 1 3t e e e e 2 At 2 2 2 e = 3 t 3 3t 1 t 3 3t e + e e + e 2 2 2 2 t 0 y (t ) = C e A( t ) (1)d 1 0 1 ( t ) 1 3( t ) t e 2e 2 y (t ) = C d 1 3 0 e ( t ) + e 3( t ) 2 2 1 ( t ) 1 3( t ) e e 6 y (t ) = C 2 1 ( t ) 3 3( t ) + e e 6 2
t
}
}
} }
1 1 1 t 1 3t 2 6 2 e 6 e y (t ) = [3 0.2] . 1 3 1 t 3 3t + e + e 2 6 2 6
163
y( t ) 1
3 exp ( t ) 2 7 exp ( t ) 5
1 exp ( 3 t ) 2 2 exp ( 3 t ) 5
1 exp ( t ) 10
3 exp ( 3 t ) 30
} }
Multiply matrices
y( t ) 1
Simplify
y( t )
d x dt
0 3
1 4
.x
0 1
.u
y = [3 0.2] .x
State space realization for this problem and the circuit below
Z( s ) L R2
u( t ) B
[
y( t )
B [
3
R1
H( s ) s
.2 s
2
4s
164
Definitions
Time Domain Solution
d x A .x dt x( t ) e
At
B .u t x( 0) 0
y Cx
}
) ) .B .u ( ) d ) ) .B .u ( ) d
exp ( A .( t t 0
} } }
State Solution
At y( t ) C .e x ( 0 )
C.
exp ( A .( t
1
Output solution
exp ( A .t ) =
1 ( sI A)
B .U( s ) ( sI
1 A) .B .U( s ) 1 A) .B .U( s )
X( s ) ( sI Y( s ) C ( sI
A)
x( 0)
C ( sI
} } } } }
State Space equation Take Laplace Transform Simplify Simplify to obtain State Solution Multiply by "c" to obtain the output solution
Y( s ) H( s ) U( s )
Y( s ) H( s ) C ( sI U( s )
1 A) .B
}
Since
So given a state space representaion, we can use the above definition to determine the poles and zeros of the system from
H (s) =
Cadj( sI A) B sI A
M 1 =
adj ( M ) M
sI
0 C .adj( sI A) B 0
165
More Definitions
# A state variable realization (A,B,C) can be found from the differential equation or a transfer function # The transfer function of the realization satisfies 1. 2. 3.
H( s )
Rational in s
H( s )
poly poly H( s ) s s
Strictly proper
2 3
for example
1 3
# A realization is minimal if it has the smallest number of states possible # A realization is minimal if H(s) has no pole-zero cancellation
# A realization is controllable if the n x n matrix has full rank (i.e., a non zero determinant)
[B
AB L An1B
# If a realization is controllable then the modes or poles of the system can be arbitrarily changed . n 1
CA
..... CA C
# If a realization is observable then the modes or poles can be observed at the output y(t) # If a realization is controllable and observable, then it is minimal # If a realization is minimal, then it is controllable and observable
166
Circuit Example
d x dt
0 3
1 4
.x
0 1
.u
x( 0)
0 0 L
Z( s ) R2
x ( 3 0.2 ) .x
Find the open loop transfer function and determine if the system is controllable and observable
u( t ) B
[
y( t )
B [
R1
H( s ) C .( sI s 3
A) 4 1
1.
B C.
s 3 s
1 4
.B
By definition
H( s ) C .
1 0.2 s 3 0.2 s 3 . B C . 1 s ( s) s ( s ) s 2 4 s 3 ( s 1 ) .( s 3 ) 3
( s) s
4s
B CA C H( s )
AB
0 1
1 4
} } }
non zero determinant means the matrix has full rank; hence the system is controllable non zero determinant means the matrix has full rank, hence the system is observable
If we have H(s) the minimality test can use to determine controllability and observability as opposed to the above matrix check
167
Given an input u(t), y(t) will have a shape that is dependent on u(t) and H(s). The shape of y(t) may not be the desired shape so we change H(s) by using "output" feedback
u ( t ) K .( r ( t ) r( t)
y( t ) )
}
K
r( t)
B
u H( s )
y( t )
}
}
Closed loop denominator
( s) 1
K .H( s )
Poles can be changed by adjusting K (see Root Locus Design) As we have seen using frequency domain techniques, we are limited in how we can change the closed-loop system by using output feedback. That is we cannot place poles arbitrarily. That is, the desired closed loop poles may not lie on the root locus, or we may not be able to obtain the exact response that is desired. Given a state-space formulation for H(s), that is controllable, we can arbitrarily place the poles using "state" feedback; hence, we can achieve any desired shape for the output response
d x Ax dt y Cx u r
+ -
Bu
}
}
& x
+
Kx u
r
+
x
1/s
A K
168
Pole Placement
State Space Representation of
H( s ) C .( sI
1 A) .B
d x A .x dt
Let
B .u Kx
y Cx r( t)
u( t )
r( t)
is a reference input
d x A .x dt d x (A dt y C .x sI .X( s )
Extra Step
B ( K .x BK ) .x
r)
substitute u(t)
Br
}
}
x( 0) ( A
BK ) .X( s )
B .R ( s )
X( s ) ( sI Y( s ) C .( sI R( s )
(A (A
1 BK ) ) .x ( 0 ) 1 B .K) ) .B
( sI
(A
1 B .K) ) .B .R ( s )
Definition of Inverse
If the system is controllable, the gain vector "K" can be calculated to place the poles by using the following formula
K ( 0 , 0 , ...... , 0 , 1 ) d ( A) d ( s )
1 ( A , B) d ( A)
Ackermans formula
s A
}
B
d ( s )
( A , B)
B , AB , ...... , A
1.
det
( A , B)
169
d x dt
0 3
1 4
.x
0 1 s
.u
y ( 3 0.2 ) .x
H( s )
0.2 .s 3 ( s 1 ) .( s 3 )
1, 3
Kx
}
s
2
d ( s ) ( s d ( A) A d ( A)
2
5)
10 .s 25 .I 0 3 0 1 1 4 1 4
25
}
Find
n 2
desired closed loop poles at s = -5
10 .A 1 4 .
d ( A) 1 4 25 0 0 25 22 18 6 2
0 3
10 .
0 3
d ( A)
( A , B ) ( B AB ) K (0 1 )
Check the answer
1 ( A , B ) . d ( A) ( 22 6 )
1
Ackermans Formula
Y( s ) C ( sI R( s ) A BK 0 3
(A 1 4
BK ) ) 0 1
1
Closed-loop TF formula
.( 22 6 ) .B
0 25
1 10
}
A
Find
BK
s 1 Y( s ) C. R( s ) 25 s 10
Substitute for
BK A BK
s 10 1 0 Y( s ) . . 1 C. R( s ) 25 s 1 ( s) ( s) s
2
10 s
25 ( s
5)
1 Y( s ) . 1 ( 3 0.2 ) . R( s ) s ( s) Y( s ) 0.2 s 3 R( s ) ( s 5 ) 2
Simplify
170
L1
B
R2
B
Co
u( t )
R1 y( t )
L2
x ( t ) B 1
z1 ( t )
Find
& = Ax + Bu x y = Cx
z2 ( t )
Define the current throught the inductor as a state Define the voltage across the capacitor as a state z1 , z2 are auxiliary variables that will be eliminated
z1 Co . x2 z1 u L.
d x dt 1 z2
} Capacitive Law
=>
z2 x2 d x dt 2
Co .
d x dt 1
d x dt 2
R1 .z2 L . R1 .z2
R1 . x2 d x dt 1
Co .
d x dt 1
}
Co .
L1
x1 L.
R2 .z1 d x dt 2
R2 Co .
R1 . x2 Co . R1
d x dt 1
L2
R1 .x2
R1 .Co .
d x u dt 1 0 1 R1 R1
d R2 . x1 dt 1 0
R1 .x2
x1 0
} Simplify
x x1 x2
R1 C Co . R1 d x dt 1 d x dt 2 R1 C Co . R1 R2 R2
L 0
d x dt 1 d x dt 2 L R2
1
x1 x2
.u
R1 Co Co . R1
0 1
R1 R1
x1 x2
R1 Co Co . R1 R2
L 0
1 0
.u
L 0 R2
0 Co . R1 R2
L R1 Co
.1
LCo . R1
171
1 d x dt Co . R1 R1 L . R1 R2 R2
R1 Co . R1 L . R1 R2 0 .x 1 .u L R1 .R2 R2
Simplify to obtain
& = Ax + Bu x
form
x2 ( t )
L1
B
R2
B
Co
u( t )
R1 y( t )
L2
x ( t ) B 1
z1 ( t )
z2 ( t ) d x dt 1 1
Co .
}
R2 .x .x
1
R1 Co . R1 R2
Co . R1 R1 .R2
.x
R2
R1
,
R2
}
R2 .u
=
y Cx
form
Let
3.8 H , 0 0.26
Co
.066 F
d x dt
.x
y ( 0.75 0.75 ) .x
172
d x dt
.x
0 0.26
.u
y ( 0.75 0.75 ) .x
}
0.2 0.2
2
H( s ) C .( sI s
1 A) .B
}
11.36 s 0.2
( sI
A)
3.79 0.2
11.36 s 4s
.1 3.79 3
Matrix Algebra
(s H( s ) C .
0.2 .( 11.36 ) s . 0
} Matrix Algebra
H( s )
3 3
4 .s
The same circuit gives two different state space forms but retains the same TF RCF SSF Physics-based SSF
1 0 0 &= x x + u 3 4 1 y = [3 0.2]x
State x does not have any physical meaning, it is just a mathematical creation
x1 x2
is the voltage across the capacitor and is the current through the inductor )
173
y = [0.75 0.75]x
For the circuit, let
}
}
u
2
Kx
d ( s ) ( s d ( A) ( A
5)
}
2
n 2 A
2
5 I)
10 .A n 2
1
25 .I
Find
d ( A)
controllability matrix
. ( A) d 2.83 r
k1 k2 .x2
k2 23.95
Controller
answer
OpAmp summer
B
x2
Current Sensor
B
R2 B
C B x1
Voltage Sensor
R1 y(t)
r(t)
B
u(t)
k1
and
k2
174
175
Observers
Problem: What if the state x is not measurable ?
Given A,B,C, y(t), u(t), and build an observer for estimating x
+ +
& x
1/s
x C
A
+
& x
1/s +
x
C
y ~ y
Observer
& = Ax + Bu x
y Cx
Y( s ) H( s ) U( s )
Plant
The observer is a device, we build to manufacture The observer gain L is used to make
x x
quickly; L is an n x 1 vector
& = Ax + L~ x y + Bu ~ = Cx Cx = C~ y = yy x
& = Ax + LC~ x x + Bu
}
}
~ x = xx
Substitute for
}
~ y
observation error
~ x
goes to zero
176
~ & = ( A LC) ~ x x
Error Dynamics
~ ~ sX ( s) ~ x (0) = ( A LC) X ( s)
~ ( sI ( A LC )) X ( s ) = ~ x (0) }
~ X ( s) = ( sI ( A LC)) 1 ~ x (0)
adj( sI ( A LC )) ~ ~ X (s) = x ( 0) sI ( A LC )
~ X ( s)
}
0
Definition of Inverse
0 ( s )
If the poles of quickly i.e.,
sI 0 ( s )
(A
LC)
~ x 0
x x
The observer gain "L" is utilized to place the observer poles in any location as long as the open loop is observable.
CA ( A , C)
1 is full-rank i.e.,
..... ..... C
det( ( A , C ) ) 0
1 L o ( A)
1
( A , C)
0 .. 0
} Ackermans formula
o ( A) = o (s) s = A
where
o ( s )
177
Example
1 0 0 &= x x + u 4 3 1 y = [3 0.2]x
} y = Cx
& = Ax + Bu x
Find an observer so
x x
in 0.5 seconds
& = Ax + L~ x y + Bu = Cx y
o ( s ) ( s ( A , C) 10 ) CA C
1 2
}
s
2
observer structure
20 .s
100
} construct
0 ( s )
}
4.321 20.19
o ( A) .
check answer
~ & = ( A LC ) ~ x x
s
~ 0 x = xx
12.962 .136 s 7.038 8.72
2
sI
(A
LC)
64.571
} } }
o ( s ) ( s o ( s ) s
2
12.962 ) .( s 20 .s
7.038 ) 10 )
Take Determinant
100 ( s
Simplify
x x
in 0.5 seconds
178
& = Ax + L~ x y + Bu
= Cx y +r u = Kx
~ y = yy
r(t) is the reference input
A, B, C are found by determining a state-space realization for the plant H(s). The vectors K and L are found by using the given design specifications and Ackermans formula.
Plant r( t)
+ -
u( t ) H( s )
y( t )
& x
1/s +
x
C
B K Controller
~ y
A L Observer
~ y = yy
179
& = Ax + Bu x y = Cx & = Ax + B( r Kx ) x y = Cx
& = Ax + L~ x y + Bu = Cx y
State-space realization for the open loop plant substitute for u(t) to obtain closed loop plant
+r u = Kx
& = Ax + B( r Kx ) + LC ( x x ) x = Cx y
Find the Error Dynamics by using
~ y = C~ x
~ x = xx
~ x
= x~ (x x)
}
BK A
If
~ x =0 & = ( A BK ) x + Br x y = Cx
x z = ~ x
Acl
A 0
BK
LC
Bcl
B 0
Ccl ( C 0 )
}
Acl
Bcl
Closed-loop TF
180
Y( s ) Ccl . sI R( s ) sI Acl
1
Acl sI
1.
Bcl
}
sI
Acl
A 0
BK A
1
BK LC
Bcl
B 0
Ccl ( C 0 ) BK (A LC)
(A 0
BK )
Plug in
Acl
( sI ( A BK )) 1 ( sI Acl ) 1 = 0
M N
Matrix Identity 1
BK ( sI ( A BK )) 1 ( sI ( A LC )) 1 ( sI ( A LC )) 1
M 0
1
} Matrix Identity
NP
1
( sI ( A BK )) 1 Y ( s) = [ C 0] R( s ) 0
Y( s ) C ( sI R( s ) (A BK ) )
1
BK ( sI ( A BK )) 1 ( sI ( A LC )) 1 B 0 ( sI ( A LC )) 1
Ccl .
Y( s ) ( C .adj( sI ( A BK ) ) ) B R( s ) sI ( A BK ) ( C .adj( sI sI (A (A BK ) BK ) ) ) B 0 0
Matrix Identity
The closed-loop TF has "2n" poles; "n" poles come from the observer and "n" poles come from the closed-loop plant. How do you know this is true ? L sets the location of the "n" observer poles. K sets the location of the "n" closed-loop plant poles. During the above calculation, the "n" observer poles given by
sI
(A
LC)
This cancellation of the observer poles is always true and leads to what is called the separation principle. Separation Principle - the gains K and L can be designed separately to achieve the desired closed-loop plant and observer pole locations, respectively.
181
Example
open loop plant
H( s ) s
.2 s
2
3 4
u( t ) H( s )
y( t )
3 .s
Design a regulator so the desired closed-loop poles are at s = -1, -2, and the observer poles are at s = -10, -12. Find a SSR using RCF
d x dt
0 4
1 3
0 1
A n 2
0 4
1 3
0 1
C ( 3 0.2 )
y ( 3 .2 ) x
Formulate Observer
& = Ax + Bu + L~ x y = Cx y
L o ( A) o ( s ) ( s ( A , C)
1
}
1 0
observer
( A , C) .
}
2
Ackermans Formula
10 ) ( s CA C
22 s
120
}
0.2
observability matrix
L (A
10 I) ( A
12 I)
.8 2.4 3 s
1 0 0.8 s
4.6 26.1
} }
answer
Check
sI
(A
LC)
13.78 82.34
Use formula
8.22
2
o ( s ) ( s
13.78 ) .( s
8.22 )
82 .( 0.08 ) s
22 .s
120
checks
Note: observer poles are placed s-10 times further out in the LHP
182
Formulate Controller
+r u = Kx
K
=
controller 1
(0 1 ) .
( A , B ) cl ( A) 2) s 0 1
1 2
}
2
Ackermans fromula
cl ( s ) ( s
1) ( s
3 .s
( A , B ) ( B AB )
1 3
}
I) .( A s 2 s 1 3 2
Controllability Matrix
K (0 1 ) .
0 1 (A
1 3
(A
2 .I ) ( 2 0 )
answer
Check
sI
BK )
2
}
}
use formula
cl ( s ) s .( s
3)
2 s
3s
checks
Review
& = Ax + L~ x y + Bu
= Cx y +r u = Kx
State-space regulator How do you find A,B,C,K and L for a given problem ?
183
State Transformation
d x Ax dt y Cx
Let T be an n x n constant nonsingular matrix i.e.
Bu
xn T
xn
x T .xn T.
d d x T . xn dt dt Bu
1 T .B .u
Premultiply by
d x A .x dt n n n y C .x
n n
Bn .u
1 An T .A .T
1 Bn T .B
Cn C .T
Definition
Two state space realizations are equivalent if 1 An T .A .T 1 Bn T .B
Cn
C .T
184
Theorem
Equivalent systems have the same transfer function Proof:
H( s ) C .( sI Hn ( s ) Cn . sI
1 A) .B
}
1.
By Definition By Definition
An
Bn
1
Hn ( s ) C .T . sI Hn ( s ) CT . sT
1
1 T .A .T
.T 1 .B
1
}
1
Substitute for
An , Bn , Cn
IT
AT
1
.T
1
} }
1 Hn ( s ) CT . T .( sI 1 Hn ( s ) C .T .T .( sI
A) T
.T
Rearrange
1 1 A) .T .T .B
} }
( MNP )
1 1 P .N .M
Hn ( s ) C .( sI
1 A) .B H( s )
Proof is complete
Theorem
Given two state space representations for the same system, the transformation T for 1 xn T x can be calculated by either method
( A , B) 1 ( A , C)
1 An , Bn An , Cn
} }
185
xn2 ( t ) u( t )
R2
B
R1 B y( t )
Co B xn1 ( t )
The same circuit gives two different state space forms but retains the same TF RCF SSF Physics-based SSF
1 0 0 &= x x + u 3 4 1 y = [3 0.2]x
xn T
( A , B ) ( B AB )
0 1
1 4 0
An , Bn ( A , B) .
Bn An .Bn
1
2.954
An , Bn
1.286 3.846
answer
Hence, the state that did not have physical meaning in the RCF form are actually linear combinations of the states in the Physics-based SSF
x T .xn
x1 x2
0.339
xn1 xn2
1.286 3.846
By Definition
3.846 .x
}
n2
State Relationships
186
d x A .x dt y c .x d x A .x dt n J n yn CJ .xn
B .u
} }
BJ .u
CJ C .T
xn T
" T " is the matrix which consists of " n " eigenvectors of the matrix "A". Put the generalized (i.e., for repeated roots) eigenvectors in " T " in the reverse order while keeping them next to each other
Theorem
If the eigenvalues of "A" are distinct then "A" is called simple. However, repeated eigenvalues does not imply that A is not simple
Theorem
"A" is simple if and only if
AJ
(i.e., there must be only zeros appearing off the main diagonal)
Definition
The eigenvalues of
AJ
AJ AJ
BJ CJ
AJ AJ
BJ CJ
is non zero. is observable if the first element of the corresponding Jordan block in
b) eigenvalue in
is non zero.
187
AJ
Never isolate a "one" on the superdiagonal - any time you have repeated eigenvalues you have the possibility of a "one" showing up on the super diagonal Draw as many lines as possible Extend lines through
BJ
and
CJ
Example - Find the JNF and discuss Controllability and Observability & = Ax + Bu x y = Cx
1 0 A= 6 5 0 B= 1 C =[ 1 1]
1 = 3 , 2 = 2
1 1 v1 = , v2 = 3 2 1 1 2 1 T 1 = T = 1 3 2 3 3 0 AJ = T 1 AT = 0 2 1 BJ = T B = 1 C J = CT = [ 2 1]
1
} } }
3 0 1 &n = x x n + u 0 2 1 y = [ 2 1]xn
188
AJ
Never isolate a "one" on the superdiagonal - any time you have repeated eigenvalues you have the possibility of a "one" showing up on the super diagonal Draw as many lines as possible Extend lines through
BJ
and
CJ
Jordan Block
1 3 0 &n = x xn + u 1 0 2 y = [ 2 1]xn
}
Jordan Block
Analysis Since "A" is a simple matrix (i.e., "A" is a main diagonal matrix) a) eigenvalue in b) eigenvalue in
AJ AJ
BJ CJ
Eigenvalue -3 -2
189
} } }
, 0 1 0
1 v 2 = 0 , v 3 0 5 1 3 T 1 = 0 1 0
5 = 3 1 0 5 1 3 0 1
eigenvectors of "A"
1 1 0 AJ = T AT = 0 1 0 0 0 2 4 BJ = T B = 0 1 C J = CT = [ 1 1 8]
1
4 1 1 0 & n = 0 1 0 x n + 0 u x 0 0 2 1 y =[ 1 1 8]x n
190
AJ
Never isolate a "one" on the superdiagonal - any time you have repeated eigenvalues you have the possibility of a "one" showing up on the super diagonal Draw as many lines as possible Extend lines through Jordan Block
BJ
and
CJ
1 & n = 0 x 0 y= [ 1
Analysis
1 0 4 1 0 xn + 0 u 0 2 1 1 8]xn
}
Jordan Block
AJ AJ
BJ CJ
b) eigenvalue in
is non zero.
Eigenvalue 1 2
Controllable No Yes
191
& = Ax + Bu x y = Cx 1 1 2 A = 0 1 3 0 0 1 1 = 1 , m = 3 0 v1 = 0 1 3 T = 0 0 1 B = 0 0 C = [1 1 0]
}
} } }
eigenvectors of "A"
C J = CT = [3 5 0]
0.33 1 BJ = T B = 0 0
}
}
192
AJ
Never isolate a "one" on the superdiagonal - any time you have repeated eigenvalues you have the possibility of a "one" showing up on the super diagonal Draw as many lines as possible Extend lines through Jordan Block
BJ
and
CJ
1 & n = 0 x 0 y = [3
Analysis
1 0 0.33 1 1 xn + 0 u 0 0 1 5 0]xn
Jordan Block
AJ AJ
BJ CJ
b) eigenvalue in
is non zero.
Eigenvalue 1
Controllable No
Observable Yes
193
AJ
Never isolate a "one" on the superdiagonal - any time you have repeated eigenvalues you have the possibility of a "one" showing up on the super diagonal Draw as many lines as possible Extend lines through
BJ
and
CJ
0 0 0 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 0 1 1B 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 &n = x x n + u 0 0 0 0 0 2 1 0 0 0 0 0 0 2 1 0 0 0 0 2 0 0 0 0 0 2 0 0 0 0 0 0 0 3 0 1 y = [1 0 1 1 2 0 0 0]xn
194
Jordan Blocks
0 0 0 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 0 1 1B 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 &n = x x n + u 0 0 0 0 0 2 1 0 0 0 0 0 0 2 1 0 0 0 0 2 0 0 0 0 0 2 0 0 0 0 0 0 0 3 0 1 y = [1 0 1 1 2 0 0 0]xn
Jordan Blocks
Analysis a) eigenvalue in
AJ AJ
BJ CJ
b) eigenvalue in
is non zero.
Eigenvalue -1 -1 -2 -3
A
Controllable No Yes
Yes Yes
Yes No
195
Polynomial Regulation
Given an open loop system
U( s )
H( s )
Y( s )
where
H( s )
B( s ) A( s )
} }
H( s ) V( s )
Reference Input
T( s ) R( s )
U( s)
B( s ) A( s )
Y( s )
Polynomial Regulator
N( s ) R( s )
such that
Y( s ) T ( s ) . V( s ) R( s )
B( s ) A( s ) N( s ) . B( s ) R( s ) A( s )
Hd ( s )
Hd ( s )
Q( s ) P( s )
} }
Restrictions 1) 2)
degree ( P ( s ) ) degree ( A( s ) )
Q(s) must contain the roots found in B ( s )
196
Step 1
Determine A(s), B(s), Q(s) and P(s) from the problem statement and check restriction 1 B( s ) open-loop TF
H( s )
A( s ) Q( s ) P( s )
} }
Hd ( s )
Restriction 1:
desired closed-loop TF
degree ( P ( s ) ) degree ( A( s ) )
and
Step 2
Find
B + ( s)
B ( s)
B( s ) = B + ( s ) B ( s ) B ( s) B + ( s)
By Definition
denotes the polynomial which contains the roots of B(s) that have positive real parts denotes the polynomial which contains the roots of B(s) that have nonpositive real parts
B( s ) = s( s + 1)( s 1) B + ( s ) = s( s + 1) B ( s ) = ( s 1)
Restriction 2: Step 3
Example
B ( s)
T ( s) =
T1 ( s )Q ( s ) B (s)
R( s ) = R1 ( s) B + ( s)
are unknown at this point.
By Definition
R1 ( s ) , T1 ( s )
Step 4
By definition
197
Step 5
T1 ( s ) , R1 ( s ) , N( s )
T1 ( s ) ( s R1 ( s ) s
L c
a)
}
1
rL
.s L
1
...... ......
r0 n0 ri , ni ri , ni
} }
N( s ) nc s
nc
.s c
At this point the coefficients denoted by Step 6 Solve the Diophantine Equation to determine
coefficients in
R1 ( s )
and
N( s )
A( s) R1 ( s) + B ( s) N ( s) = P( s)T1 ( s)
Diophantine Equation
After matching the coefficients in the above equation, we obtain the polynomials Step 7 Construct R(s) and T(s)
R1 ( s ) , N ( s )
T ( s) =
T1 ( s )Q ( s ) B (s)
R( s ) = R1 ( s) B + ( s)
198
Example
V( s )
Reference Input
T( s ) R( s )
Y( s ) H( s )
-
Polynomial Regulator
N( s ) R( s )
Given
H( s )
s s
2
Y( s ) s 1 V( s ) ( s 1 ) 2
Step 1
Determine A(s), B(s), Q(s) and P(s) from the problem statement and check restriction 1
B( s ) s 1 2 A( s ) s Q( s ) s 1 Hd ( s ) P( s ) ( s 1) 2 H( s )
open-loop TF
desired closed-loop TF
degree ( P ( s ) ) degree ( A( s ) ) 2
OK
B + ( s)
and
B ( s)
H( s )
B( s ) s 1 2 A( s ) s B( s ) = ( s 1)
Plant
B ( s ) = ( s 1) B + (s) = 1
Restrictions 2)
} B( s ) = B + ( s ) B ( s ) } roots of B(s) with positive real parts } roots of B(s) with nonpositive real parts
Q( s ) ( s
Step 3
1)
B ( s ) = ( s 1)
OK
T ( s) =
T1 ( s )Q ( s ) T1 ( s )( s 1) = B ( s) ( s 1)
R( s) = R1 ( s) B + ( s ) = R1 ( s)(1)
R1 ( s ) , T1 ( s )
199
Step 4
L 1
C 1
T1 ( s ) , R1 ( s ) , N( s )
T1
Rule of Thumb
T1 ( s ) ( s R1 ( s ) s N( s ) n1 s
Step 6
10 ) r0 n0
(s
10 )
} }
L 1
deg{R1} = 1
} deg{N} = 1
A( s) R1 ( s) + B ( s) N ( s) = P( s)T1 ( s)
s s s s n0 N( s )
Step 7 2 3 2
}
s
2
ri , ni coefficients in R1 ( s )
and
N( s )
Diophantine Equation
s
2
r0
(s n1 s
2 2
1 ) n1 s n1
n0 n0 s
2s
3
1 (s 12 s
2
10 ) 21 s
}
10
Insert Polynomial
s r0 r0 n1 n1 n0 10
n0 s
Simplify
s ( 12 ) s ( 21 ) n1 31 n0 10 r0 43 R1 ( s ) s
}
43
Simplify
} }
( 31 s
10 )
R1 ( s )
and
N( s )
T ( s) =
T1 ( s )Q ( s ) ( s 1) = ( s + 10) = s + 10 B ( s) ( s 1)
R( s) = R1 ( s) B + ( s) = R1 ( s)(1) = s + 43
200
Plant
V( s )
s s
10 43
s s
2
Y( s )
( 31 s 10 ) s 43
Check your answer
V( s )
s s
10 43
(s s
3
43 ) .( s
2
1) 10
Y( s )
12 s
21 s
V( s ) s
3
(s
10 ) .( s
2
1) 10
Y( s )
12 s
21 s
} } }
V( s ) (s
(s
10 ) .( s
2
1) 2s 1
Y( s )
10 ) . s
Factor denominator
V( s )
(s (s
1) 1)
2
Y( s )
checks since
Y( s ) s 1 V( s ) ( s 1 ) 2
201
Nonlinear Systems
d x f ( x , u) dt d x dt
=
f ( x , u ) is an n x 1 vector
T
General form for most nonlinear systems Linearized about an equlibrium point
f x
f x + u x xe
x xe
}
T
xe . d x Ax dt Bu
d x dt
Ax + Bu
f A= x
x = xe
f B= u
x = xe
where
f n f1 f 2 x x x 1 1 1 f 1 f 2 f n x x x 2 2 2 f = x f1 f 2 f n x n x n x n
Calculation for
f x
f f f1 f 2 n = u u u u
The equilibrium point
Calculation for
f u
xe
xe
202
Example
m L g
Dynamic Equation
- input control torque - mass of robot - link of robot - link angle - gravitational coefficient
Linearize the system about the equilibrium point obtain a state-space realization.
e = 0 &e = 0
and then
& = Ax + Bu x y = Cx
( i.e., find A, B, C)
x x = 1 = & x2
& = f ( x, u ) x
form.
original dynamics
&& = g sin( ) + 1 u } rearrange mL && x1 = From &2 = x x2 = & problem } &1 = x2 x statement
x2 f1 & x = f ( x, u ) = = 1 sin( ) g x + u f 1 2 mL
& = f ( x, u ) x
n=2
form
203
Step 2)
Find
f x
and
f u
f1 f 2 x1 x1 f = x f1 f 2 x x 2 2
f f1 = u u
f 2 u
u
}
}
x2 f & = f ( x, u ) = 1 = x f 2 g (sin( x1 )) +
f x
0 g .cos x1 1 0
1 mL
}
}
substitute for
f1 and f2
f = 0 u
1 ml
substitute for
f1 and f2
x1e e 0 x2e
& =
=0
T
x xe = 1e = &e x2 e e 0 1 = + g 0
0 = 1 mL
f A= x
f B= u
x = xe
x = xe
Step 4)
y = y = x1
y (1 0 ) . C (1 0 ) x1 x2
Since
x1
by definition
204
m L g
0 1 0 &= x x+ u g 0 1 / mL y = [1 0]x
u
Find the open-loop TF and the open loop impulse response. Then assuming are measurable, design a state feedback controller which places the poles of the closed loop system at -1.
,&
Y( s ) H( s ) C ( sI U( s ) ( sI A)
2 1
1 A) .B
}
Matrix Algebra
s g g
1 s
. 1 g s ( s)
s 1
( s) s
s 1 0 1 1 / mL . H ( s ) = [1 0]. = 2 g s 1 / mL ( s ) s g
Y( s ) H( s )
if
Open-loop TF
U( s ) 1
Impulse response
Y( s ) s
1 mL g . s g
}
}
gt
Factor denominator
1 1 2 g 2 g 1 Y ( s) = + mL s + g s g
y( t ) 1 2 g mL y( t ) ( t )
Link Falls Down
PFE Tool
. e
gt
}
t
205
Control Design
}
r
State Equation
Kx 1)
2
so poles are at s = -1
Objective
d ( s ) ( s d ( A)
}
1 0 0 1 1 mL 0
1
n 2
2
0 1 g 0 0
1 g
2 1
2g g 1 g 2 1
Let
s A
K (0 1 ) .
1 mL
2g
Ackermans formula
K = [mL( g + 1) 2mL]
Check the answer
} }
answer
( s)
sI
(A
BK )
1 0 1 0 0 [ ] A BK = mL g mL ( 1 ) 2 + = 1 2 g 0 1 / mL
( s) ( s) s
2
Find A - BK
s 0 0 s 2s
0 1 1 (s
1 2 1)
2
s 1 s
1 2
checks
Kx
where
x = &
so
206
r( t)
+ -
u(t)
Encoder
y( t )
mL( g
+
1)
&
2 mL
Numerical Differentiator
Computer
Control
207
Same Example
m L g
What if
&
Solution:
Polynomial Regulator
V( s )
Reference Input
T( s ) R( s )
Y( s ) H( s )
-
Polynomial Regulator
N( s ) R( s )
Given
H( s )
1 mL s
2
}
g
Open-loop TF
Y( s ) 1 V( s ) ( s 1 ) 2
Step 1
Determine A(s), B(s), Q(s) and P(s) from the problem statement and check restriction 1
H( s )
B( s ) 1 A( s ) mL . s 2 Q( s ) 1 P( s ) ( s 1) 2
} }
open-loop TF
Hd ( s )
desired closed-loop TF
Restrictions 1)
degree ( P ( s ) ) degree ( A( s ) ) 2
OK
208
Step 2
Find
B + ( s)
and
B ( s)
1 B( s ) A( s ) mL . s 2 1 B( s ) = mL 1 B + (s) = mL B ( s) = 1 H( s )
Restrictions 2)
Plant
Q( s ) 1
Step 3
B ( s) = 1
OK
T ( s) =
T1 ( s )Q ( s ) T1 ( s ).1 = 1 B (s)
R( s ) = R1 ( s ) B + ( s ) =
R1 ( s ) mL
R1 ( s ) , T1 ( s )
Step 4
L 1
C 1
T1 ( s ) , R1 ( s ) , N( s )
T1
Rule of Thumb
T1 ( s ) ( s R1 ( s ) s N( s ) n1 s r0 , n1 , n0
10 ) r0 n0
(s
10 )
} }
L 1
deg{R1} = 1
} deg{N} = 1
209
Step 6
A( s) R1 ( s) + B ( s) N ( s) = P( s)T1 ( s)
s s
3 2
ri , ni coefficients in R1 ( s ) 1 .( s 21 s
and
N( s )
g . s r0 .s
2
ro g .s
( 1 ) . n1 .s g .r0 n1 .s
n0 n0 s
3
2s 12 s
2
10 ) 10
} }
Insert Polynomial
Simplify
2 s . r0
2 s .( 12 )
r0 12 n1 g n0 10 21 g .r0 10 g .12
s. g g .r0 r0 12
n1
s .( 21 )
n0 10 n1 g
21
n0 12 g
10
R1 ( s ) s N( s ) ( g
Step 7
12 21 ) s ( 12 g 10 )
}
s + 12 mL
Gives
R1 ( s )
and
N( s )
T ( s) =
T1 ( s )Q ( s ) ( s + 10)(1) = = s + 10 B ( s) (1)
R( s ) = R1 ( s ) B + ( s ) =
210
Plant
V( s )
(s s
10 ) 12 mL
1 mL . s
2
Y( s ) g
(g
21 ) s s
( 12 g 12 mL
10 )
Block Diagram for closed-loop system
V( s )
(s s
10 ) 12 mL
(s (s 12 ) . s
2
12 ) . ( 21
1 mL g) s ( 12 g 10 )
Y( s )
V( s ) s
3
s 12 s
2
10 21 s 10
Y( s )
V( s ) (s
(s 10 ) . s
10 )
2
Y( s ) 1
2s
} }
Factor denominator
V( s ) (s
1 1)
2
Y( s )
checks since
Y( s ) 1 V( s ) ( s 1 ) 2
211
Appendix
Causality
We also note that even though only causal systems can be implemented in real time, noncausal systems are still of inetrest. For one thing, the independent variable may not always be time, as for example, in optics. Second, even when time is the independent variable, noncausal systems are useful in providing benchmarks of performance; the ideal band-pass filter is a classic example. Furthermore, as the reader may know, noncausal systems may often be satisfactorily approximated by causal systems plus a delay. FInally, noncausal systems may arise in the course of analysis, when decomposing or recombining causal systems.
- x(0+)
L+ [ (t )] = s.s 1 1 = 0 = L+ [0]
(t )] L+ [
L-
0+
(t )e
st
dt
=0
The expected formula for the Laplace transforms of from the use of the formulas. =
(t )
which also agrees with the direct evaluation. More detailed discussions of the
(t )] L -[
=L
[d1(t ) / dt ]
s.s 1 0 = 1
C. Desoer, Linear Systems --- A State-Space Approach, MGraw-Hill, New York, 1963.
"Derivatives" of Discontinuous Functions and Delta Functions. The introduction of the delta function enables us to give meaning to the derivative of a function at a point of discontinuity. We should note that Q (.) can be regarded as the derivative of a step function. This 0 and the use derivative relationship is certainly true (in the conventional sense) as long as of the symbol (.) allows us to extend it to the case where = 0.
(t ) = lim p (t ) = lim
0
Having gotten this far, we can go further and, using the same ideas, introduce a derivative for the delta function and in fact also second-and higher-order derivatives. To introduce the ( 1) ( . ) , we write derivative, say
(1) (t ) = d (t ) / dt = lim
(t ) (t ) 0
212
, the
function in Fig a) tends towards the unit impulse (.), while the function in Fig b) approaches what can be interpreted as its derivative. Since for very small the functionin Fig b) approaches what can be interpreted as its derivative. Snce for very small the function in Fig b) consists of two narrow but strong pulses of opposite sign, the function in the limit is called the unit doublet. It is an idealization of a dipole in electromagnetic theory and a couple (torque) in mechanics. ( 2) 3 This procedure can be similarly extended to define ( . ) , ( . ) and so on: We merely with smoother and smoother approximating sequences for (.). The derivatives obtained by the above procedures may be called generalized derivatives since the conventional derivatives are not defined.
a)
b)
1 1
c) d)
213