E C E - 4 0 9

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

Example of a System that needs Control

Motor Turning a Mechanical Load T J T : torque input X : output position J : inertia B : friction b Newton's Law

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

after some algebraic manipulation

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)

Block Diagram of the Pendubot Setup

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)

is Newton's law of motion

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 :

means the nth derivative)

Time Invariant
The differential equations will be assumed to be time-invariant. That is, the coefficients, denoted by

ai and bi

do not depend explicitly on time.

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

Given a system with two differential inputs and

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( )

for all values of

The ''D" Notation for Analyzing DE's

Examine the differential equation

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

So the above equation can be written as n n 1. D y an 1 D y .......

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

characteristic equation for 2

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

Determine the "shape" of y(t) eigen values or

Roots of

( D)

are sometimes referred to as the poles

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

Time for a Movie

Click on image to start movie

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

Definition of the Laplace Transform (one - sided)

f(t), s

denotes the Laplace Transform Operation

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

If f(t) is single valued for Apply the vertical line test

( f ( t ) ) .e 0
p

0 t

dt <

then

f( t)

is Laplace transformable for Re(s) > 0 Checking the region of convergence t f ( t ) e .u 1

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

This is sometimes called the

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

Some Properties of Laplace Transforms

1.

( f ( t ) ) F( s )

d f( t) dt t 0

sF ( s ) f ( ) d

f 0

Often used for solving DEs

2.

F( s ) s f 0
p

3. Initial Value Theorem:

lim t 0
p

f( t) s

lim

sF ( s )

4. Final Value Theorem:

f( ) t

lim

f( t)

lim sF ( s ) s 0

often used for setting performance specs for control design

5.

t f a ( f( t e
at .

a .F ( a .s )
sT

} } }

scaling ("a" is a constant)

6. 7.

T) ) e f( t)

F( s ) a)

delay ("T" is a constant) frequency scaling ("a" is a constant)

F( s

8.

f1 ( t ) * f2 ( t ) f1 ( t ) * f2 ( t )

F1 ( s ) .F2 ( s ) t

convolution convolution equation is avoided by multiplication in the frequency domain

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)

Solution of Differential Equations by Laplace Transform

1. Take the Laplace Transform of both sides of the equation to work in the frequency domain 2. Solve the algebraic equation for the unknown variable 3. Take the Inverse Laplace Transform to find the time domain expression Example Find y(t)

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

Partial Fraction Expansion

Y( s ) Y( s )

4 4 e
4t

Preparing the expression for using the tables

Inverse Laplace Transform

y( t )

2e

The answer

Check if the initial condition is satisfied

y( 0 ) 3
3t 4t

A quick good way to catch errors

Plug in the solution and check if the original equation is recovered

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

} } }

The answer is correct

Finding the Output from a Linear Differential Equation

Given a system and an input, how do we find the output ?

u
System

We use an nth order differential equation to describe the system as follows

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 )

We could now substitute for

U( s )

and then use the Inverse Laplace Transform to find

y( t )

Side Note: Transfer Function

Y( s ) H( s ) U( s )

bn s
n

. sn 1

1 1

..... ......

b1 . s a1 . s

b0 a0

an

. sn

} }

Definition of Transfer Function

A transfer function is an Input/Output Relation

Remember : Set the initial conditions to zero when calculating the transfer function

10

More Math Tools

The Partial Fraction tool is important for taking the Inverse Laplace Transform

Example for distinct roots

H( s )

s4

8s s
3

25 s 6s
2

31 s 6

15

Problem : Find h(t)

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

Now work with

H1 ( s ) via the PFE tool.

Step 2:

Find

h1 ( t ) s
3

by using the PFE tool and tables

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

There is no easy way to factor the denominator.

K1 K2

and

K3

11

Residue Formula:

Ki

si .H( s )

s si
2

This can be used as long as the root at si is not repeated

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)

How do you check this ? Get a common denominator

(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)

PFE for Repeated Roots

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
.....

repeated root multiplicity "q"

s1 .... sr - distinct roots ...... s

with

H( s )

K1 s K1 s1
.....

....

Kr s
with

Aq sn
q

sr

s A1

Find

Kr

Find

Aq

the residue formula

the formula below

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

Example for repeated roots

Find h(t) for

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

Take the Inverse Laplace Transform with the tables

h( t ) 1

10 .exp ( t )

11 .t .exp ( t )

answer

13

More Math Tools

( x ( t ) ) X( s )
Property 9.

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)

is defined this way p skeleton" "The 0

f ( t ) . ( t

T) dt f ( T)

Definition of delta function

Transfer Function
Input / Output approach to system modeling

u( t ) H( s ) Y( s ) H( s ) U( s ) y
n

y( t )

Transfer function is calculated by assuming all initial conditions are zero (n 1)

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

H(s) is stable if all the poles have negative real parts

14

Example
+

Vc

Given +

Vc

voltage across the capacitor

1F i y 1
_

u
_

u : input voltage y : output voltage i : current t u 2e Find y(t)

Vc 0 y i

1 d V dt c
=>

Voltage loop Take Derivative

{ { {

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

Take Laplace Transform of DE

( 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

from problem (Note

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

Substitute initial conditions Solve for Y(s)

}
1 s

Y( s )

}
1

Apply Residue Formula for Distinct Roots =2

d (s ds

2 1) .

2s (s 1)
t 2

s
t

}
t

Use Repeated Root Formula

Inverse Laplace Transform from Table

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

are positive constants

at

s (s a)
2

Side note on the Impulse Response

Y( s ) u( t ) ( t ) H( s )

h ( t ) is the Impulse Response for

for

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 )

Example: Calculating an Output

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

From the tables

y( t ) 3 U( s ) s 3 s s
2

poles always come in conjugate pairs if the coefficients are real

Y ( s ) H ( s ) .U ( s )

2s

Use quadratic formula for 2nd order roots

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

Find A,B and C

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

After equating the coefficients in the above equation

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

still does not look quite like the tables

1 2
2

s (s 1)

1
2

2s

5 (s

(s

1 2 (s

2 1)
2

complete the squares

which looks like

s a)
2

a 0
2

0 (s a)
2

From the tables......... ( s

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

Take the inverse Laplace Transform to obtain answer

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

How do you check the answer ?

d y( 0 ) 0 dt 2 d y 2. Plug the solution into 2 dt

1. Check

y( 0 )

d y dt

5 .y u ( t )

u( t ) 3

18

Checking for Stability Without Factoring the Denominator H( s ) N( s ) D( s )

n

D( s ) s n an
1

an

.......

a1 s

a0

Denominator determines stability

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

Form the continued fraction expansion

If all

h1 .......hn are positive then all of the roots of D( s )

are in the OLHP (stable)

( Note OLHP - open left half plane-poles have negative real parts)

Example: Continued Fraction Expansion

Is

D( s )
3

stable ?

D( s ) s D1 ( s ) s
3

6s

12 s

Denominator determines stability 2

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

Continued Fraction Expansion is completed

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)

So all the roots of

D( s )

20

Example: Continued Fraction Expansion

Find the values of K such that the system described by the following equation is stable(Note K>0)

( 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 >0 14 0 < K< 784

=>

For stability K should satisfy For

K> 784

=>

there is at least one unstable pole

21

Describing a Second Order System

d 2 y 2 . n . d y dt n s
2 2

d t2 U

2.

Y( s ) H( s ) U( s ) n undamped natural frequency damping ratio

Y n
2

2 . . n .s . n

The poles of

H( s )

are

s s

j . n . 1
and

and

. n

j . n . 1

which is also written as

j . d

j . d

is called the time constant

d j

is the damped natural frequency axis

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

> 1 1 0< < 1 0

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

More Laplace Transforms

Second Order System

If

t .e

n .t

1
=

n n

} }

Poles are real and repeated

If

sin n .t

Poles are purely imaginary

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

Stability: Time Domain and Frequency Domain Relationships

Second Definition of Stability
A system is stable if its impulse response approaches zero as time goes to infinity, that is 1( H( s ) ) 0 lim h ( t ) =

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

So stability is only dependent on the poles of the transfer function

24

Time for a Movie

Click on image to start movie

25

Block Diagram Simplification

U(s)

H(s)

Y(s)

Goal of Block Diagram Simplification

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)

Most common Block Diagram

} } } } }

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)

Block Diagram has been simplified

The need for simplification of Block Diagrams spawned the creation of the rules and procedures

27

Block Diagram Simplification Rules

Cascade Y P1 . P2 . X X X P1 P1 . P2 P2 Y

simplified

Parallel X

P1

P2 . X P1 P2 + + Y

P1

P2

simplified

Remove a Block from a Path X P2 P1 P2 Y

}
Y

equivalent to above block. 1 X

P1 Y P2 . P2 Feedback loop Y P1 X P2 . Y P1

P2 . X

+ -

P1 P2

X 1

P1 P1 . P2

simplified

Remove a Block from a Path

X 1 P2

+ -

Y P1 . P2

Y=
equivalent to above block.

1 P 1P 2 X P2 1 + P 1P 2

28

Moving a Summing Point ahead of a Block or beyond + X P + Z Y

+ + 1 P

Z Y

Z P. X

X Y +

X Y

P P

+ +

Z P( X

Y)

Moving a Take-off point Y P 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

(Some problems have multiple inputs)

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

Response of C(s) due to R(s) only

CR = Set R = 0

G1 . G2 1 G1 . G2

.R u +

}
G2 G2

Block Diagram Simplification Rules

G1

} }

Response of C(s) due to U(s) only

+ + 1

Cu

Rearrange Block Diagram

G1 G2 1 G1 . G2

Cu The total response is given by C Cu CR =

}
Block Diagram Simplification Rules G2 G1 . G2 U

G1 . G2 1 G1 . G2

.R

30

Rule-Based Reduction of Complicated Diagrams

1. Combine Cascade and Parallel Blocks 2. Eliminate Feedback Loops.

(Method 1)

Example 1

G3 + R + + + G1 G4 G2 + C

H1

}
G2 G3 C

Find C ( s) R ( s)

H2

+ -

+ +

G1 G4

H1

Combined Cascade and Parallel Blocks

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

Simplified Feedback Path

Simplified Feedback Path

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

identify parallel paths

+ -

H ( s)

} } } }
add parallel paths close feedback loop

add parallel paths

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

Time for a Movie

Click on image to start movie

34

Signal Flow Graphs ( Method 2)

Nomenclature

m Y Y m .x b

mathematical expression

branch - which way you go Node - variables are represented by nodes

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

Example and Definitions

A42 A21 X1 X2 A23

Definition - A path is a continuous unidirectional succession of branches along which no node is passed more than once.

A32

A33 X3 A43 X4

Signal Flow Graph that is used to explain Masons terminology

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

X1 .X2 .X4 A21 A42 } X2 .X3 .X2 A32 A23 }

path gain for a forward path

Definition - A loop gain is the product of the branch gains of the loop Example -loop gain for a loop

Construction of signal flow graphs +

R
+

C
Block Diagram

C
Signal Flow Graph

37

Simplification of Signal Flow Graphs

SFG

} First draw the signal flow graph

C P =H = i i R i Pi Pjk
= ith forward path

Masons Rule

= jth possible gain product of K non-touching loops

= 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

= evaluated with all loops touching

Pi

being eliminated

Definitions - Two loops, paths, or a loop and a path are non-touching if they have no nodes in common

38

Example - Simplification of an SFG

1

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

- Simplify Block Diagram

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

2) Find the forward paths

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

} = 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.........) - ......

i 1

= evaluated with all loops touching 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

same as before but with different technique

41

Example -Simplify Block Diagram

G2
+

R
+ +

G1 H1

1) First, draw the Signal Flow Graph

G2 G1
1 1

R
1 1

2) Find the forward Paths

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

Example - Putting it all together

If R(s) = 1/s Find c(t) 1 + +

R( s )
+ -

1/s

1/s 1 1

C( s )

1 1) Draw SFG

1/s 1 1

1/s

2) Find the forward loops

P1

R
1 1/s 1 1 1/s 1 1

3) Find the loops

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

Inverse Laplace Transform

44

Example Simplify SFG

b1 u
1 1/s 1/s 1/s

b2 b3

a2 a1 a3

Reachable Canonical Form

1) Find the forward paths

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

2) Find the Loops

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

General Form of a 3rd order Transfer Function

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

Alternative method to solve

R( s )
1

G1 G1 H1
1

G2 G2 H2

Y( s )

closed both loops

Y( s ) R( s )

= 1

G1 G2 G1 H1 G2 H2 G1 G2 H1 H2

multiply cascade blocks

48

Time for a Movie

Click on image to start movie

49

Root Locus: Construction and Design

Plant

R( s )

 B

Gp ( s )

Y( s ) K is a positive parameter that can be changed

K .Gp ( s ) Y( s ) R ( s ) 1 K .Gp ( s ) s Gp ( s )
m n

}
......

Closed Loop Transfer Function

am bn

1 1

.s m
1

.s n

....... n

a0 N ( s ) b D( s )
0

General Form for the Transfer Function

N( s )

and D ( s ) are polynomials where m

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)

in the s-plane change as

A "Locus" of these roots plotted in the s-plane as a function of K is called the

Different ways to write the same thing

( s) 1

K.

N( s ) D( s ) D( s )

K .N ( s ) 0

Gives the roots of the closed loop system

50

Root Locus Rules

1. Starting points (K = 0) 2. Termination points 3. Number of distinct root loci The root loci start at the open-loop poles. The root loci terminate at the open-loop zeros. The open loop zeros include those at infinity. There will be as many root loci as the largest number of finite open loop poles or zeros. For the majority of systems, the number of finite open-loop poles will be greater than the number of finite open-loop zeros. The root loci are symmetrical with respect to the real axis. The asymptotes intersect the real axis at a point given by centroid formula 6. Root Locus locations on the real axis 7. Rool locus asymptotes

4. Symmetry of root loci 5. Asymptote intersection

i =

open loop poles open loop zeros nm

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

Asymptote Chart in the s plane

120

90

72

RD 1

RD 2

RD 3 x

RD 4
in the asymptote charts above

RD 5

The centroids are marked

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

Open loop poles are at

s 0, 2, 4

Closed Loop Transfer Function

4)

Locus Form } Root for the Denominator Step 2. Real Axis

Step 1. Pole Zero Plot

j j

; ;
-4 -2

; ;
-4 -2

Step 3. Centroid / Asymptotes

centroid

centroid RD 3

2 3

-4

; ;
-2

Step 4. Draw the Root Locus

4

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 )

Open loop zero Open loop poles

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

Closed Loop Transfer Function

( s) 1

K.

(s

1 ) .( s

3j )

Root Locus Form for the Denominator

Step 1. Pole Zero Plot

Step 2. Real Axis

;
|
-3 -2 -1

;
|
centroid

R ;

R ;

-3 -2 -1

Step 3. Centroid / Asymptotes

;
5 2

Centroid

1 2

( 2)

R ;
-3 -2 -1

RD 2

10 8 6

Step 4. Draw the Root Locus

Imag Axis

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)

Open loop poles

1, 3

Y( s ) R( s )

1 ( s 1 ) .( s 3 ) 1 K. ( s 1 ) .( s 3 ) 1 1 ) .( s

}
}

Closed Loop Transfer Function

( s) 1

K.

(s

3)

Root Locus Form for the Denominator

Step 1. Pole Zero Plot

Step 2. Real Axis

j
-3

j
centroid

; ;

-1

-3

; ;

-1

Step 3. Centroid / Asymptotes

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

-0.5 -1 -1.5 -2 -2.5 -4 -3 -2 -1 0 1 2

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 )

0 0 1 2 3 Tim e (se c.) 4 5 6

Closed Loop System



Y( s )
B

R( s )
Let R(s) = unit step

Gp ( s )

S tep Response 1.2

K 30 K 10 K 4 K 1

0.8 Amplitude

0.6

}
10

as K changes it is clear that the transient response can be "shaped"

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 )

1 s .( s 3 ) .( s 3 j .7.4 ) .( s 3 j .7.4 ) 1 K. s .( s 3 ) .( s 3 j .7.4 ) .( s 3 j .7.4 ) 1 j .7.4 ) .( s

Closed-Loop Transfer Function

( s) 1

K.

s .( s

3 ) .( s

j .7.4 )

Locus Form } Root for the Denominator

Step 1. Pole Zero Plot

Step 2. Real Axis

; ;
-3 ;

-3

; ;

centroid

Step 3. Centroid Asymptotes

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

Magnitude and Angle Conditions

K .Gp ( s ) Y( s ) R ( s ) 1 K .Gp ( s )
Poles are determined by

R( s )

 B

Y( s ) K Gp ( s )

( s) 1 ( s)

K .Gp ( s ) s clp

characteristic equation

=0

clp denotes closed-loop pole

Magnitude Condition

( s) 1

K .Gp ( s ) K .Gp ( s )

s clp

=0

=>

K .Gp ( s )

= -1

s clp

Expression for finding clps

s clp

} }

Take the magnitude of both sides (Note K is a real number)

K . Gp ( s )

s clp

=1

Magnitude Condition

Angle Condition

K .Gp ( s )

= -1

}
=

Expression for finding clps

s clp Angle ( 1 )

Angle K .Gp ( s )

Take the angle of both sides

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

Find the value of K which places a closed loop pole at -5

R( s ) Gp ( s ) (s s 2 1 ) .( s 4)

 B

Y( s ) K Gp ( s )

Find the transfer function

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

Draw the Root Locus

-0.2 -0.4 -0.6 -0.8 -1 -6 -5 -4 -3 -2 -1 0 1 2

Note that -5 lies on the root locus

Real Axis

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)

5 Angle ( 4 ) 180 ( 180 180 ) 180

= - 180

Angle ( 3 )

( Angle ( 1 ) )

Condition } Angle is satisfied

59

Design Example: DC Motor

The differential equation for a dc motor is given by

d t2
where

( t)

Ra La

B .d ( t) J dt ( t)

Ra B . La J

Ga .If La .J

. ( t )

Gaf .If J .La

.V ( t ) a

Va ( t )

is the input voltage and

is the motor speed

After testing the motor and determining the parameters, we obtained

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

Poles of the plant are given by

6 .s

Open Loop system that we want to control

1 and s
then

Let

Va ( t ) ( t )

( t)

1
=

Gp ( s ) .Va ( s )
Pole Plot
4

( s) ( t)

Imag Axis

0.25 0.25 PFE Tool s 1 s 5 Inverse 1. 1. exp ( 5 .t ) exp ( t ) Laplace Transform 4 4

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

axis (Assume that the system is stable)

The dominant pole corresponds to the slowest mode The open loop systems output response will be dominated by e t Why ?

60

Design Example - Continued

Let us examine the response to a step input

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

Inverse Laplace Transform

Amplitude

Problem: The response of the system maybe too slow to suit our specifications Solution: Add Feedback

0.12 0.1 0.08 0.06

Va ( s ) R( s ) 
B

( s) Gp ( s )

0.04 0.02 0 0 1 2 3 Time (sec.) 4 5 6

( s) R( s ) 1
Find K so that the closed loop poles are at

K .Gp ( s ) K .Gp ( s )

Closed Loop Transfer Function

Draw the Root Locus

3 1 1 ) .( s

j .3

and

j .3
4

From Magnitude Condition

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

places the closed loop poles at

-2 -3 -4 -6

j .3

and

j .3

-5

-4

-3

-2

-1

Real Axis

61

Design Example - Continued

With K = 13 and R(s) = 1/s , find ( t )

( 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

Closed Loop Transfer Function

6s

18 1 s 18 1 3
2

Use PFE Tool

( s ) 13 .

1 .1 18 s

(s

3)

13 .

1 .1 18 s

1. s 3 3 18 ( s 3 ) 2 3 2

Make it look like the tables

( 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

Inverse Laplace Transform

Closed Loop

Open Loop

Faster Dominant Pole

Slower Dominant Pole

Step Response

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 2 4 Time (sec.) 6 8 10

Closed Loop Note it takes

Open Loop

Amplitude

5 seconds to die out 5t while it takes e one second to die out

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)

Gc ( s ) K j .0.5 and s j .0.5

Find K so the closed loop poles are at

0.5

0.5

K Y( s ) s .( s 1 ) K 2 R( s ) K s s 1 . s ( s 1)

Closed Loop Transfer Function

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

Draw the Root Locus

-1

-1.5 -2

-1.5

-1

-0.5

0.5

Real Axis

Method 1: Coefficient Matching

( 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

After equating coefficients, we obtain

63

Method 2: Magnitude Condition

( 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

Applications of FVT and IVT for open loop systems

Gp ( s ) r( t) 1 R( s ) s Y( s )
2 s .( s

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

Open Loop System

Step Response

y( t ) A

Bt

C .e

Use PFE tool

}
2 Time (sec.) 3 4 5

y( t )

Initial Value Theorem

f 0
p

lim s

s .F ( s )

y 0

lim s

s
2 s .( s

0 1)

} }

Checks with graph

Final Value Theorem

f( ) lim s .F ( s ) s 0 y( ) s lim s 0 s (s
2.

1)

Checks with graph

64

Application of FVT and IVT for Closed Loop System

R( s ) R( s ) 1 s 0.5 s .( s

1 1)

Y( s )

Closed Loop System

1. 1 Y( s ) 2 s .( s 1 ) 0.5 2 R( s ) 1. 1 s 0.5 s 1 . 2 s ( s 1) 0.5 s. s

2

closed loop TF

Y( s )

0.5 0.5 s. (s 0.5 ) C .e

2

s
0.5 .t .

( 0.5 )

}
}

Use PFE tool Inverse Laplace Transform

y( t ) A

Be

cos ( 0.5 .t )

0.5 .t .

sin ( 0.5 .t )

FVT :
y( ) s lim 0 s . s2

0.5 .s s 0.5 .s 0.5

Step Response

1 0.9 0.8

IVT :
y( 0 ) s lim
2 s. s

0.7 Amplitude 0.6 0.5 0.4 0.3 0.2 0.1

0 0.5

} y( t )
0 5 10 Time (sec.) 15 20

FVT - tells us A = 1 IVT - tells us B = -1

65

2nd Order Time Domain Transient Specifications

H( s ) s n
2 2

2 . n .s

standard form for 2nd order Transfer Function

1. Percent Overshoot (PO) max value of output - final value of output PO = final value of output
100 90 80

time domain calculation

PO 100 .exp 1

.
2

70 60 50 40 30 20 10 0 0 0.2 0.4 0.6 0.8 1

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

Typical Output to a Step Input

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

Time Domain Specifications

Conversion Formulas

Frequency Domain Specifications

67

Steady State Error (SSE) Specification

+

R( s )
-

Gp ( s )

Y( s )

e( t) r( t)

y( t )

SSE t

lim
r(t)

e( t)

lim sE ( s ) s 0

Final Value Theorem

} SSE
y(t)

Amplitude

Time (sec)

Example - Calculation of SSE

E( s ) R( s )

Y( s ) K .Gp ( s ) 1 K .Gp ( s ) K .Gp ( s ) 1

Y( s ) .R ( s )

K .Gp ( s ) 1 K .Gp ( s )

.R ( s )

} From Block Diagram

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

were available, then SSE could be calculated numerically

E( s )

1 .R ( s ) K .Gp ( s )

Simplify

SSE

lim s . 1 s 0

1 .R ( s ) K .Gp ( s )

Final Value Theorem

68

Example - Calculation of SSE

+

R( s )
-

Gp ( s )

Y( s )

H( s ) K .Gp ( s ) 1 K .Gp ( s ) .H( s )

E( s ) R( s )

Y( s )

Y( s ) K .Gp ( s )

.R ( s )

From Block Diagram

E( s ) R( s ) . 1

K .Gp ( s ) .H( s ) 1) .R ( s )

substitute Y(s)

1 E( s )

K .Gp ( s ) .( H( s ) 1 K .Gp ( s ) .H( s ) 1

common denominator

SSE

lim s . s 0

K .Gp ( s ) ( H( s ) 1
and

1)

K .Gp ( s ) H( s ) R( s )

.R ( s )

Final Value Theorem

If

K , Gp ( s ) , H( s )

were available, then the SSE could be calculated

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

(Settling Time: 2% criterion)

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

Formulate desired closed loop characteristic equation

dcl ( s ) s

2 . . n .s

6 .s s 3

18 j3
and

Desired closed loop poles are are

4 3 2

j3

} quadratic formula

Imag Axis

1 0

}
-5 -4 -3 -2 -1 0 1 2

Root locus for

-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

Find K by Magnitude Condition

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

2) Find y(t) when R(s)

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

} From Block Diagram K=13

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)

Make it look like the tables

y( t ) 13 .

cos ( 3 .t )

1. e 18

sin ( 3 .t )

Inverse Laplace Transform

Interesting check on PFE Calculation

y( )

lim s .Y( s ) s 0 lim s

13 18

} Final Value Theorem

y( 0 )

s .Y( s ) 0

} Initial Value Theorem

71

3) Find SSE when R(s) = 1/s

E( s ) R( s )

Y( s )

}
.R ( s )

Definition

Y( s )

K .Gp ( s ) 1 K .Gp ( s )

}
.R ( s )

From Block Diagram

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

Lead Compensator (Using Magnitude and Angle Condition)

+

R( s )
-

Gc ( s )

1 s .( s 1)

Y( s )

Find Let

Gc ( s )

such that the dominant closed loop poles are at

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

Root Locus does not go through

3 j3

-2

-3 -3.5

Real A xis

If

Gc ( s ) K . K. b

(s (s (s (s

a) b)

Lead Compensation =>

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

Root Locus goes through

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

are more dominant that "-c"

How do we find the values for a, b and K which ensure that are the closed-loop poles ?

3 j3

73

Find Closed Loop Transfer Function

Y( s ) R( s )

K( s a ) s .( s b ) .( s 1 ) s a 1 K. ( s b ) .s .( s 1 )

From the Block Diagram

( s) 1

K.

( s a) . s ( s b ) .( s

=0

Achieves the desired result

1)

j3

Angle Condition: Find "a" and "b"

Angle K .

s .( s

( s a) b ) .( s

1)

j .3

= -180

} }

One Equation Two Unknowns

Angle Conditions

Angle

(s (s

a) . 1 . b) s ( s 1)

3
=

j .3 180

= - 180

One Equation Two Unknowns

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

=>

b 18 Angle ( AB ) Angle ( A) Angle ( B )

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

Magnitude Condition - Find "K"

K.

(s

s 3 18 ) .( s ) .( s

1)

j .3 1

K.

15

j .3 j .3 . 3 j .3 . 2

j .3

K.

3 1 ( 15.3 ) .( 4.24 ) .( 3.65 ) s 3 s 18

K 79 ( approximately)
answer

Gc ( s ) 79 .

75

Lead Compensator (Using Coefficient Matching)

+

R( s )
-

Gc ( s )

Gp ( s )

Y( s )

Find

Gc ( s ) 1 s .( s

so that the dominant closed loop poles are at

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

Select a Lead Compensator for the same reason as before

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)

actual closed loop denominator

j
Imag Axis

-3+j3 0 -3-j3

-b

-c

-a

Still need to draw the root locus to understand the problem

Real Axis

cl ( s ) s

(b

1 ) .s

(b

K) .s 3

a .K j .3 )

} }

actual closed loop denominator desired closed loop denominator

dcl ( s ) ( s dcl ( s ) ( s dcl ( s ) s

3

c ) .( s c) . s (6
2

j .3 ) .( s 18 ( 18

6 .s
2

}
6 c) s 18 c

simplification

c) s

76

Equate the desired denominator to the actual denominator

dcl ( s ) cl ( s ) 6 18 c b 1 K

6c b

aK 18 c

}
18 K K 78

3 equations, 4 unknowns (cannot solve) Let

a 3

(same reason as before)

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

checks with previous answer

Note we have discovered the third closed loop pole in the process (i.e., s = - 13)

Interesting Side Note

Previous equation can be used to determine how big a can get

c b

b 6c

18

18 . c a

eliminate K and b

c 6 .c

18

18 . c a

one equation and two unknowns

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 )

write this immediately

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 )

final value theorem

A 1

Closed Loop Step Response

78

Example

Find the range of K which ensures closed loop stability

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

Closed-loop Transfer Function

cl ( s ) 1

K .P ( s ) .H( s ) 1

K.

s 3 . s 0.1 ( s

s 1 2 ) .( s

4)

Denominator in Root Locus Form

Draw the Root Locus

1.5

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

0.1 ) .( s 0.1 ) . s 6.1 s ( 6.1

2 2

2 ) .( s 6s 8.6 s

4) 8 0.8 ( 8.6

K .( s K. s Ks
2 2

Denominator in Polynomial Form

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

desired denominator (see root locus)

a s

simplified

After equating the coefficients, we obtain 1) 2) 3)

6.1 8.6 0.8

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

K) .( 8.6 ( 8.6 18.8 K

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

roots is the answer

1.5

Since K cannot be negative

K 1.94

For

0 < K< 1.94

, the system is stable

1

Imag Axis

0.5

-0.5

-1

-1.5 -6

-4

-2

Real Axis

80

Lead Compensator: Design Example

R( s )
+ -

Gc ( s )

4 s .( s 2)

Y( s )

Design

Gc ( s )

so that the dominant closed loop poles are at and

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 )

Closed Loop Transfer Function

cl ( s ) 1

K.

s s

a . 4 . b s ( s 2)

Imag Axis

j2 3

-b

-c

-a

j2 3

Draw the Root Locus

Real Axis

81

Method I Angle/Magnitude Condition

Angle Condition Let

a 2.9 s 2.9 . s b (s

Do you know why ?

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

Method II: Coefficient Matching

K. 1 a) 4 . b) s ( s 2) s a . 4 s .( s K. s b s( s 2) (2 s
3

Y( s ) R( s )

(s (s

4 K (s b ) .( s 2 )

a) ( Ks

Ka ) 4

Closed-Loop Transfer Function

cl ( s ) s dcl ( s )

b ) .s

(2 b 2

4 K) .s

4 Ka

}
c)

simplified actual denominator

2 (4

j .2 3 . s c) s
2

j .2 3 .( s 4 .c ) .s 16 .c

desired denominator (see root locus)

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

a 2.9 1 2 0 0 4 11.6 3.4 5.4 4.68 c . b K

Do you know why ?

1 4 16 c b K

3 equations 3 unknowns

Use calculator to solve (same answer as before)

Note: "c" is between "a" and "b"

Imag Axis

j2 3

-b

-c

-a

j2 3

Real Axis

83

Steady State Error Specification

Decrease the steady state error so that the output goes close to the input ( Remember that e(t) = r(t) - y(t) )

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 )

Closed Loop Transfer Function

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

Apply the final value theorem

R( s )

1 s

so as K goes up SSE goes down

84

Same Problem

Y( s ) R( s )

K. 1

1 ( s 1 ) .( s 5 ) 1 K. ( s 1 ) .( s 5 ) K 1 ) .( s

Find K so that the closed loop poles are at

3 j3

cl ( s ) 1
4 3 2

(s

5)

Root Locus Form

Imag Axis

1 0

-1 -2 -3 -4 -6

Draw the Root Locus

-5

-4

-3

-2

-1

Real Axis

Magnitude Condition (Find K)

1 ( s + 1)( s + 5)
1 1

=1
s =3+3 j

=>

K 13

SSE

5 K 18 5

From the previous slide with K = 13

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

It would be desireable to place the pole and

reduce the SSE at the same time

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

To meet a transient specification

Gc ( s ) E( s )

c 0.1

d 0.01

1 . K Gp ( s ) .Gc ( s ) lim s .E ( s ) s 0 lim s 0 1 13 .

}
1

From block diagram

E( s ) R( s )

Y( s )

SSE ( s )

1 lim s . . s 0 s 1

1 K .Gp ( s ) .Gc ( s )

Final Value Theorem

SSE ( s )

s 0.1 . s 0.01 ( s

1 1 ) .( s

}
5)

Plug in the transfer functions

SSE ( s )

1 1 130 27 1 5
and

( approximately)

From previous slide

If

Gc ( s ) 1
with

K 13

SSE

1 ( approximately) 3

Gc ( s )

s 0.1 s 0.01

SSE is a factor of 7.5 better than with Gc (s) = 1

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

while neglecting the

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)

Denominator form for Root Locus

25 20 15

Imag Axis

10 5 0 -5

j5 -b -c -a - j5

-10 -15 -20 -25 -15 -10 -5 0 5

Draw Root Locus

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

Lag Compensator Design:

Gc ( s ) 56.6 . s s 10 . s 18 s
Lead

c d
Lag

} General Form for the Lead/Lag

1 .R ( s ) 0.01

SSE

lim s . s 0 1

G c( s) G p( s)

} From Block Diagram

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

obtained a ratio for "c" and "d"

"c" is typically selected to be 0.1; hence, d = 0.00625

Gc ( s ) 56.6 .

s s

10 . s 0.1 18.3 s 0.00625

Final Compensator Design

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

10 . s 0.1 . 18.3 s .00625 ( s

1 1 ) .( s

5)

= 0

j5

Exact Characteristic Equation

Angle Condition
s s 10 . s 0.1 . 18.3 .00625 ( s 1 1 ) .( s
should be equal to -180;

5)

j5

however, it is equal to -179.798

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 )

Denominator form for Root Locus

15

10

Imag Axis

-b -a
0

j2 - j2

-c
-5

Draw Root Locus

-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

Let a = 3.5 Why ?

b = 4.38

Magnitude Condition
s (s 4.38 ) ( s 3.5 1) ( s 3 ) .( s 10 )
s=-2+2j

=1

=>

K 50

Lag Compensator Design:

Gc ( s ) 50.0 . s s 3.5 . s 4.38 s c d

General Form for the Lead/Lag

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

} From Block Diagram

SSE =

1 ( 50 ) ( 3.5 ) ( c ) 30 ( 4.38 ) d
= 74.32

= 0.01

obtained a ratio for "c" and "d"

"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

3.5 . s 0.1 . 4.38 s .001345 ( s

1 1 ) .( s 3 ) .( s 10 ) s 2 j2

= 0

..... Exact Characteristic Equation

91

Angle Condition
s
angle

3.5 . s 0.1 . 4.38 s .001345 ( s

1 1 ) .( s 3 ) .( s 10 ) s 2 j2

should be equal to -180;

however, it is equal to -182.398

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

Lead Compensator Only

output response to a step input Note the similarities of the transient responses

Lead/Lag Compensator Lead Compensator Only

output response to a step input

Note the differences of the steady state responses

93

Time for a Movie

Click on image to start movie

94

Bode Plot Construction

H( s ) K . (s s n .( s a) ( s c ) .( s s j j a
n j ( j ) . c

b) d ) .( s e)

Typical Transfer Function

a ,b ,c ,d ,e

are

real positive numbers Bode Form : Let and normalize

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 )

and then convert to dB i.e.,

20 .log( H( j ) )

Case 2: n > 0 Calculate the - intercept

20 .n

intercept
dB / dec dB / dec

Kab cde

1 n

Calculate the initial slope

Final Magnitude Slope = -20 . RD

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

a c ( Angle ( j ( Angle ( j a) Angle ( j Angle ( j c) ) 0

= 0 degrees

Initial Angle =

Final Angle =

a)

c) )

= 90 - 90 = 0 degrees

1 0.8 0.6

Imag Axis

0.4 0.2 0

-0.2 -0.4 -0.6 -0.8 -1 -10 -8 -6 -4 -2 0

Pole / Zero Plot

Real Axis

96

Example - Bode Plot Construction

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

( use semilog paper )

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

-0.2 -0.4 -0.6 -0.8 -1

10

}
0

Pole / Zero Plot

-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

Example Bode Plot Construction

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

Initial slope = -20.n dB / dec = -20 dB / dec

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

-20 dB/dec -20 dB/dec

}
102

( use semilog paper )

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

Pole / Zero Plot

-0 .2 -0 .4 -0 .6 -0 .8 -1

100

10

R e a l A x is

1) Initial Phase

Angle ( H( j ) )

0 j 1 Angle ( j ) Angle j 1 1 Angle j 100 1 0

10 = 0 90 0 0

Angle

= 90
Final Angle

Angle ( H( j ) ) j 10

1 Angle ( j ) Angle j 1 1 Angle j 100 1

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

0 Phase (deg); Magnitude (dB)

-50

-100

-100 -120 -140 -160

-2 -1 0 1 2 3

10

10

10

10

10

10

Frequency (rad/sec)

Bode Plots for Repeated Roots

H (s) =

1 s n
2

H (s) =

1 n
2

1 j n 1 j n 1

}
dB

Bode Form

Magnitude Plot 1) Initial Magnitude => n = 0 => Calculate initial magnitude

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

-20 -40 Decade

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

Pole / Zero Plot

Imag Axis

0.4 0.2 0

-0.2 -0.4 -0.6 -0.8

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

Final Phase Initial Phase

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

0 -10 Phase (deg); Magnitude (dB) -20 -30 -40

-50 -100 -150 10

-1

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

} }

Closed Loop Transfer Function Root Locus form for denominator

K .H ( s )

Imag Axis

40 20 0 -20 -40 -60 -80

j d

c 100

10

}
j d
0 20

Draw the Root Locus

-100

-120

-100

-80

-60

-40

-20

Real Axis

For

0 < K< K

, the closed-loop system is marginally stable if at K

the closed-loop poles are at Angle Condition

c , j d , j d

( see Root Locus )

Angle ( H( s ) )

= - 180

These two equations can

s j d

Magnitude Condition

}
= 1

be used to find

and

however, the angle condition equation would be difficult to solve Why ?

K . H( s )

s j d

109

Bode Solution for Finding

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

A Bode Plot is a graph of

H( j )

and

Angle ( H( j ) )

for all

; hence

d
With

can be directly read off the phase plot

d
u

, we can read off

H j d
u

directly off the magnitude plot ;

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

Let us continue with the problem of finding characterized by

for the closed loop system

( s) 1

K.

1 s 1 s 1 . 10 1 . s 100 1

110

Bode Solution for finding

0

and

Magnitude (dB)

-20 -40 -60 -80 -100 -120 0

- 40 dB

}
-180 deg
-1 0 1 2 3

Bode Plot from previous problem

Phase (deg)

-50 -100 -150 -200 -250 10 10 10

d = 20

10

10

Frequency (rad/sec)

H j d

40 dB
H j d 20

} }
}

Read from the Bode Plot

10

in dB

For u

100 0 < K< 100

100 80 60

answer

, the closed-loop is stable

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

Bode Compensator Design

R ( s )
B

Y( s ) K H( s )

Similar to the root-locus, we use the characteristic equation in the form

cl ( s ) 1

K .H ( s )

to examine the stability / performance of Bode Performance Parameters

Magnitude (dB)

KH ( j )

Phase (deg)

KH ( j p )
-180

10

-1

10

10

10

10

Frequency (rad/sec)

(gain crossover frequency) is equal to the frequency where

K .H( j)

0 dB 180

(phase crossover frequency) is equal to the frequency where Angle ( H ( j) )

Phase Margin -

pm 180
(If

Angle H j .p
unstable )

pm< 0 then the closed-loop system is K .H j .

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

Root Locus design parameters

- damping ratio - natural frequency

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

is a measure of relative stability

Im{s}

x n

cos ( )
Re{s}

Hence,

pm

is a measure of relative stability

113

Bode Design Problem

R( s )

 B

Gp ( s )

Y( s )

Gp ( s )

1 s. s 2 1

Find K so that

pm 45

degrees for the closed loop system

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

Angle Gp ( j) Angle Gp jp pm 180

Assume K = 1

3) From i.e.

, calculate the desired

From given

pm

Desired Angle Gp jp Angle Gp jp

4)

From the desired the desired

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

This formula is derived from the magnitude condition as done before

114

Step 1

( s) 1

K. s.

1 s 2 1

closed loop denominator Bode Plot form

Step 2
40 Magnitude (dB) 20 0 -20

K 1

Phase (deg)

-100 -120 -140 -160 10

-2

}
10
-1

Draw Bode Plot for

KGp ( s )

10

10

Frequency (rad/sec)

Step 3

Desired Angle Gp jp Desired Angle Gp jp

pm 45

180 180 135

115

Step 4 Given Desired

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

Raise or lower magnitude plot by changing K

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

desired original 0 shift 9 dB

shift 20 9 20

9 dB

; therefore

Step 7

K 10

=>

K 10

=>

K 2.8

116

Root Locus Design for the same Problem

R( s ) 
B

Y( s ) K Gp ( s )

Gp ( s )

1 s. s 2 1

Find K so that the closed loop system has a

pm 45 deg

K .Gp ( s ) Y( s ) R ( s ) 1 K .Gp ( s ) K. 1 2 s .( s

From Block Diagram

Y( s ) R( s )

2) 2 K. s .( s 2 ) 2 s .( s 2)

}
}

Closed Loop Transfer Function

( s) 1
2 1.5 1

K.

Root Locus From for the denominator

jb

Imag Axis

0.5 0

-0.5 -1

}
-jb
0 1 2

Draw Root Locus

-1.5 -2 -3

-2

-1

Real Axis

If Note:

pm 45 cos ( )

=>

0.45

From conversion formulas

117

Desired closed loop poles must have a damping ratio of

0.45

cos ( )

=>

63 deg

Formulas =>

b 1 sin ( 63 ) sin ( 27 )
Desired closed loop poles

}
2 1

Law of sines

b 2

From Root Locus

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 ?

For this problem, these three specifications are equivalent 1) 2) 3)

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

closed loop denominator Bode Plot form

Step 2
0 Magnitude (dB)

-50

For

K 1

-100

0 Phase (deg) -50 -100 -150 -200 -250 10

-1 0 1 2 3

Draw Bode Plot

10

10

10

10

Frequency (rad/sec)

Step 3

Desired Angle Gp jp Desired Angle Gp jp

pm 45

180 180 135

119

Step 4 Given Desired

Angle Gp jp

135

Find

0 Magnitude (dB)

-50

For

K 1

-100

0 Phase (deg) -50 -100 -150 -200 -250 10

-1

-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

Raise or lower magnitude plot by changing K

0 Phase (deg) -50 -100 -150 -200 -250 10

-1 0 1

-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

Time for a Movie

Click on image to start movie

121

Nyquist Plots/Nyquist Stability Criterion

R( s ) +
_

Y( s ) K Gp ( s )

Y( s ) R( s ) ( s)

KGp ( s ) 1
=

KGp ( s ) 1 KGp ( s )

Closed loop TF

Denominator Form for Nyquist Analysis

1) Plot Nyquist Plot (i.e., a Polar Plot) for

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 )

inside s-plane contour.

N denotes the net number CW encirclements of the point

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

denotes the number of closed-loop poles inside the s-plane contour.

Zc N

P0

Formula for calculation purposes

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 )}

Draw a Polar plot on a rectangular coordinate system

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

Example - Determine stability with Nyquist Plot R( s )

+ _

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.

Closed Loop Transfer Function

( s) 1

s .( s

1 1 ) .( s

10 )

the } for denominator

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

Step 2) Plot a pole/zero plot for

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.

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 )}

Nyquist Plot with K=1

r=

Im{KGp(j )}

-1

C,D

A r Re{KGp(j)}

Nyquist Plot for different values of K

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}

Nyquist Contour CW + CCW -

-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 )

inside the contour)

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

Number of closed loop poles inside the contour

small K big K

0 2

Nyquist Criterion table

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

Root Locus agrees with the Nyquist Criterion

15

10

Gp ( s )

s .( s

1 1 ) .( s

10 )

Imag Axis

}
-12 -10 -8 -6 -4 -2 0 2 4

Root Locus for

( 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

Nyquist Criterion table

128

Nyquist Performance Specification Parameters

d H(j ) -1 pm

Im{H(j )} Re{H(j )}

unit circle

-1

Nyquist Plot

pm phase margin H j . p 1

p gain crossover frequency

Phase Crossover frequency

Gain Margin =

1 H j .

1 d

Nyquist Plot contains the same type of performance specifications as the Bode Plot and the Root Locus

129

Example - Clockwise Contour R ( s )+

_

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

Closed Loop Transfer Function

1

K.

s s

1 1

the } for denominator

Nyquist Form
Imag Axis

0.8 0.6 0.4 0.2 0

s j

K .( j 1 ) K .H ( s ) ( j 1 )

-0.2

Assume K=1

-0.4 -0.6 -0.8 -1 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2

Step 2) Plot a pole/zero plot for

H ( s)

on the s-plane contour

Real Axis

picture and include measurement scheme to the

axis

Step 3) Enter a magnitude/phase table entry for each critical point.

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)

inside the contour)

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

Number of closed loop poles inside the contour

small K big K

0 1

Nyquist Criterion table

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

Same Example - Counterclockwise Contour R ( s )+

_

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

Step 2) Plot a pole/zero plot for

on the s-plane contour

picture and include measurement scheme to the

axis

Step 3) Enter a magnitude/phase table entry for each critical point.

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

Note: Numbers in the table are exactly the same as before

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)

inside the contour)

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

Number of closed loop poles inside the contour

small K big K

Nyquist Criterion table

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

Example - Clockwise Contour R ( s )+

_

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

Closed Loop Transfer Function

1 ) .( s

2) ( j ) .( j

}
K

Nyquist Form for the denominator

Step 1) Let

s j

K .H ( s )

1 ) .( j j

2)

} K=1

Assume

Step 2) Plot a pole/zero plot for

H ( s)

on the s-plane contour axis

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 )

1 j . j 1 . j 2 1) Angle( j 2) Angle( j ) Angle( 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)

inside the contour)

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

Number of closed loop poles inside the contour

small K big K

0 2

Nyquist Criterion table

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

Same Example - Counterclockwise Contour R ( s )+

_

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

Step 2) Plot a pole/zero plot for

H ( s)

on the s-plane contour

picture and include measurement scheme to the

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)

inside the contour)

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

Number of closed loop poles inside the contour

small K big K

Nyquist Criterion table

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

Example: Using Nyquist Plots to Draw Root Loci

Given the Nyquist Plot and the Transfer Function draw the Root Locus

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

The Nyquist Plot is given

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

; the number of poles inside the contour

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

Nyquist Criterion table

=>

The above root locus is wrong

138

N
small K medium K big K

Zc 2 0 2

Stability No Yes No

2 0 2

Nyquist Criterion table

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

The above result matches that of the Nyquist criterion

K1* =

1 b

and

* K2 =a

-b -1

-1 /a

139

Example: Using Compensators

Compensator Plant

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

Plant Transfer Function

K 1

( s) 1

} for the

Nyquist Form denominator

R 0

Re{s}

Step 1) Let

s j

K .Gp ( j ) Gp ( s )

K j .( j 1 ) j
axis

Assume K = 1

Step 2) Plot a pole/zero plot for

on the s-plane contour

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 "Blow up"

Im {s} B

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

180 270 180 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 )

inside the contour)

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

Number of closed loop poles inside the contour

Stability No No

small K big K

1 1

Nyquist Criterion table

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

Same Example: Different Compensator

Compensator Plant

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

Nyquist Form for the denominator

Step 1) Let

s j

K .Gp ( j ) .Gc ( j ) K . Gp ( s ) .Gc ( s ) j

( 2 j 1 ) j .( j 1 )

}
C

Assume K = 1

Step 2) Plot a pole/zero plot for

on the s-plane contour axis

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 )

j2 1 j . j 1 Angle ( j2 1) Angle ( j ) Angle ( j 1)

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)}

Nyquist Plot with K=1

Im{s} C

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 )

inside the contour)

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

Number of closed loop poles inside the contour

Stability No Yes

small K big K

1 1

Nyquist Criterion table

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

"a" can be obtained off of the Bode plot

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 )

Y( s ) K .H ( s ) Closed Loop Transfer Function . R( s ) 1 K H ( s ) s 1 Nyquist Form ( s) 1 K. for the . ( s 1 ) ( s 10 ) denominator

Step 1) Let

s j

K .H( s )

K .( j 1 ) ( j 1 ) .( j 10 ) j

} Assume K=1

Step 2) Plot a pole/zero plot for

H ( s)

on the s-plane contour axis

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

180 90 0 j 1 10 1) Angle ( j 1) Angle ( j 10 )

A -10 -1 1 C Re{s}

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)

inside the contour)

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

Number of closed loop poles inside the contour

small K big K

0 1

Nyquist Criterion table

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

the closed loop system is stable

(see Nyquist Plot)

145

Time for a Movie

Click on image to start movie

146

State Space Analysis

All of the control problems we have studied in this class have been solved in the frequency domain. We will now concentrate on developing control solutions in the time domain for the same type of systems.

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

contains the DE coefficients

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

Example Show that the differential equation

(3)

a1 . y

( 2)

a2 . y

(1)

a3 . y b1 . u

( 2)

b2 . u

(1)

b3 . u

n 3
third-order DE

can be written as follows.

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

A is a 3 x 3 matrix B is a 3 x 1 vector C is a 1 x 3 vector x is a 3 x 1 vector

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

Initial Conditions are equal to zero.

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

Multiply matrices out

From the above equations we can see that

&1 = x2 x &2 = x3 x
Therefore

& & & & & & x 1 = x3 x1 = x3

&1 + b1 & & y = b3 x1 + b2 x x 1

Direct Substitution

148

Y( s ) X1 ( s ) x1
( 3)

b1 s

b2 . s 1

b3

Take Laplace Transform

a3 . x1

a2 . x1

(1)

a1 . x1

(2)

u}

Direct Substitution

X1 ( s ) U( s ) s
3

1 a1 . s
2

a2 . s

} a3

Take Laplace Transform

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

State Space Form (RCF)

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

Example Find a state-space representation for the circuit below

Z( s ) L R2

+ u( t ) _

x +
y( t )

_ x

R1

Z(s) is the impedance to the right of the two x's

Note: Capacitor Impedance =

1 Cs

Inductor Impedance =

Ls

First get a transfer function for the circuit

Y( s ) Z( s ) U ( s ) Z ( s ) Ls Z( s )
=

}
1 Cs

Voltage Division

R1

||

R2 R1 Cs 1 Cs

Calculate Z(s) from the circuit

R1 R2 Z( s ) R1 Z( s ) R2

}
R1 1

Apply Parallel Law

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

Final Form for TF

3.8 H ,

C = .066 F

Typical Values

Y( s ) .2 . s 3 H( s ) 2 U( s ) s 4.s 3

Substitute numerical values

1 0 &= x x + 3 4

y =[ 3 0.2] x

0 1u

x x = 1 x2

answer via RCF pattern

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

for some scalar constants

A square matrix n x n has rank n if and only if

152

Eigenvalues of a Matrix

I
n

A
1

0 ...... an

} Equation for determining eigenvalues . an 0 } above expression always reduces

to finding the roots.

a1 .

Note: Eigenvalues and poles are equivalent

Procedure for determining Eigenvectors

Note: The number of eigenvectors = number of eigenvalues = "n" Case 1) All of the eigenvalues are distinct Case 2) Some eigenvalues are repeated

(A

I) .v 0

- Use generalized procedure

Case 1: Distinct Eigenvalues Example

A I

0 6 A
2

1 5 0 0 5 2

Find the eigenvalues

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

Two distinct eigenvalues

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

Case 2: Procedure for finding Eigenvectors for Repeated Eigenvalues

Step 1: For an eigenvalue repeated m times in an n dimensional matrix A, Find the smallest p such that Rank Step 2: For Find Step 3: Find a Note

m n

(A 1 k p

.I)

m }

This step gives us "p"

Nk

= Rank

(A

I )

k 1 Rank

(A

I ) }

This step gives us the

Nks

xp

such that but

(A

p I) .xp 0

(A

I )

.x

then find

xj ( A

I) .xj

1 where

(j p

1,p

2 , ....... , 1 )

} This step gives us

eigenvectors.

All the x vectors you find are generalized eigenvectors

Step 4: Reduce each If all

Nk

by 1

Nk 0

then the procedure is complete

} This step checks

for special cases.

If not continue to step 5 Step 5: Find the largest k for which Find a vector

Nk

is not zero. Call it

that is linearly independent of previous vector x. Then find 1 where

xj ( A

I) .xj

1,

2 , ...... , 1

154

Example

Find the eigenvectors of

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

Find smallest "p" so the Rank

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:

Nk Rank .( A N1 Rank .( A N2 Rank .( A N3 Rank .( A

Rank .( A I ) I ) I )

for

1 k p 3

Rank .( A Rank .( A Rank .( A (A 0 0 0

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

From above calculation

0 1 2 x1 0 0 0

2 0

3 0 0

0 0 3 . 0

0 0 3 . 3

Last two eigenvectors

Step 4: After reducing all

by one, all the results become zero. Hence we stop at this stage

x3 , x2 , x1

are the eigenvectors

156

Example

Find the eigenvectors of

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:

Find smallest "p" so the

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:

Nk Rank .( A N1 Rank .( A N2 Rank .( A

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

are the eigen vectors

158

Motivation of State-Space Solution

Scalar System

d x ax dt s .X( s ) X( s )

bu

"a" and "b" are scalar constants

Solve the equation for x(t)

x( 0) b .U( s ) s a

a .X( s ) b .U( s ) x( 0) s a

Use the Laplace Transform

Simplify

After taking the Inverse Laplace Transform

b x (t ) = e at x(0) + '1 U ( s) s a
Let H ( s ) =

Inverse Laplace Transform

b h(t ) = e at b sa

x (t ) = e at x (0) + h(t ) u(t ) x (t ) = e x (0) + h(t )u( )d

at 0 t

} } }

Convolution Property

Convolution Integral

t x( t ) e
at

x( 0) 0

exp ( a ( t

) ) .b .u ( ) d

Substitute for h(t) to obtain the solution for x(t)

Initial Condition Contribution

Input Contribution

Solution to the Matrix Equation

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

Looks same as the solution to the scalar system

is the solution of the equation where

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!

......

} }

Use Taylor Series

A .t 2!

2 2

A .t 3!

3 3

.....

Use Taylor Series

d exp ( A .t ) A .exp ( A .t ) exp ( A .t ) .A dt A

2 2 .A .t ( 1 ) .( 2 ) 3 2 3 .A .t ( 1 ) .( 2 ) .( 3 ) 4 3 4 .A .t ( 1 ) .( 2 ) .( 3 ) .( 4 )

d exp ( A .t ) 0 dt

.......

Take time derrivative

d exp ( A .t ) A I dt
Note:

A .t

A .t 2!

2 2

A .t 3!

3 3

....

A .exp ( A .t )

Factor out "A"

d exp ( At ) exp ( A .t ) .A dt

Why ?

because AB = BA if B = A

3)

We also need Leibnitz Rule

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

is the partial derivative with respect to x

160

Solution to the Matrix Equation

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

exp ( A .t ) which is called the State Transition Matrix

161

Calculating the State Transition Matrix

Note

exp ( A .t ) =

1 ( sI A)

By Definition

In the solution for x(t) we require

exp ( A .t )

hence we need a method for calculating

exp ( A .t )

'{e at } = (s a )

}
1

If "a" is a scalar

'{e At } = (sI A)
exp ( A .t )

"A" is a n x n matrix and "I" is the n x n identity matrix

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)

} }

Set up for PFE tool

1 exp ( A .t ) =

1.5 ( s 1) 1.5 ( s 1)

0.5 0.5 ( s 3) ( s 1) 1.5 0.5 ( s 3) ( s 1) 1. exp ( 3 .t ) 2 3. exp ( 3 .t ) 2

Use PFE tool

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

Take Inverse Laplace Transform

162

Example : Finding the output using the state space solution

Find y(t) if u(t) = unit step

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

}
}

State space form

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

}
}

From Previous Problem

Substitute "B" and u(t)

} }

Substitute exp(At) and multiply matrices

Integrate term by term

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

Plug limits of integration and "C"

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 )

Does this make sense in the circuit ? Why?

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

Circuit that we calculated a SSR for

H( s ) s

.2 s
2

4s

Transfer function for circuit

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

State Space equation

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)

State transition matrix

Frequency Domain Solution

d x A .x B .u dt s .I .X( s ) x ( 0 ) A .X( s ) B .U( s ) ( s .I A ) .X ( s ) x ( 0 )

1 A) .x ( 0 )

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

"Open Loop" Transfer Function

Y( s ) H( s ) U( s )

Y( s ) H( s ) C ( sI U( s )

1 A) .B

}
Since

From above output solution with x(0)=0

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

Poles are given by

sI

0 C .adj( sI A) B 0

Zeros are given by

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 )

has real coefficients

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

# A realization is observable if the n x n

..... CA C

has full rank ( i.e., a non zero determinant )

# 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

Poles of H(s) are at -1 and -3

( s) s

4s

open-loop characteristic equation

Calculate the controllability and observability matrices with n=2

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

0.6 2.2 3 0.2 0.2 .( s 15 ) ( s 1 ) .( s 3 )

The system is clearly minimal - no pole / zero cancellation

If we have H(s) the minimality test can use to determine controllability and observability as opposed to the above matrix check

167

Control Motivation from our Previous Work

u( t ) H( s ) y( t )

Open loop system

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

Control Input (Classical Control)

is the reference input

r( t)

 B

u H( s )

y( t )

}
}
Closed loop denominator

Closed loop system

( 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
+

State Space Equation

Kx u

r
+

State Feedback Law - K is 1 x n

x
1/s

A K

State Space Block Diagram

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)

State Space Form

u( t )

State Feedback Law

K is a 1 x n vector of feedback gains used to place the closed loop poles

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

}
}

Closed loop state space system

x( 0) ( A

BK ) .X( s )

B .R ( s )

Take the Laplace Transform

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 )

Solve for X(s)

Closed loop TF x(0) = 0

Y( s ) C .adj( sI ( A B .K) ) .B R( s ) sI ( A BK ) C .adj( sI sI (A (A BK ) BK ) ) .B 0 0

Definition of Inverse

Gives zeros of the closed loop TF

Gives poles of the closed-loop TF

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 )

specifies the desired location

of the closed loop poles

The open loop system will be controllable as long as the matrix

( A , B)

B , AB , ...... , A

1.

has full rank i.e.,

det

( A , B)

169

Example - State Feedback Control

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 )

Open loop poles are at Let

1, 3

Kx

}
s
2

State Feedback Controller

Find K to place the closed-loop poles at s = -5

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

n 2 Calculate the Controllability Matrix

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

Take the inverse of and substitute for B

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

answer is correct Why ?

170

A New State Space Realization for the Circuit

x2 ( t )


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 )

for our circuit

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

Kirchoffs current law use z1

d x dt 2

R1 .z2 L . R1 .z2

R1 . x2 d x dt 1

Co .

d x dt 1

}
Co .

Write equation for voltage loop

L1

x1 L.

R2 .z1 d x dt 2

R2 Co .

R1 . x2 Co . R1

d x dt 1

Write equation for voltage loop

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

Put in Matrix Form

R1 Co Co . R1

0 1

R1 R1

x1 x2

R1 Co Co . R1 R2

L 0

1 0

.u

L 0 R2

..... Matrix Algebra

0 Co . R1 R2

L R1 Co

.1

Calculate Matrix Inverse

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

Find the output of the system

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

y R1 .z2 R1 . x2 y R1 .x2 R1 .Co . R1 R1 R1

=

Co .

}
R2 .x .x
1

From the circuit

R1 Co . R1 R2

Co . R1 R1 .R2

.x

x1 Substitute for from above d t matrix equation

R2

R1
,

R2

}
R2 .u
=

y Cx

form

Let

3.8 H , 0 0.26

Co

.066 F

same numbers as before

d x dt

3.79 11.36 0.2 0.2

.x

y ( 0.75 0.75 ) .x

Final State Space form for the circuit

172

Calculate the Open Loop Transfer Function for the Circuit

d x dt

3.79 11.36 0.2 0.2

.x

0 0.26

.u

y ( 0.75 0.75 ) .x

}
0.2 0.2
2

State Space Equation

H( s ) C .( sI s

1 A) .B

}
11.36 s 0.2

Open loop TF formula 1

( sI

A)

3.79 0.2

11.36 s 4s

.1 3.79 3

Matrix Algebra

(s H( s ) C .

3.79 ) .( s s 0.2 0.2 0.2 .s s

2

0.2 ) 11.36 s 3.79

0.2 .( 11.36 ) s . 0

loop characteristic } open equation

2.95 . 1 ( 0.75 0.75 ) . .1 . 0.26 0.26 s 0.9854

} Matrix Algebra

H( s )

3 3

4 .s

same TF we obtained before

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

3.79 11.36 0 &= x x+ u 0.2 0.2 0.26 y = [0.75 0.75] x

State x does not have any physical meaning, it is just a mathematical creation

State x has physical meaning (i.e.,

x1 x2

is the voltage across the capacitor and is the current through the inductor )

These physics-based states facilitates the construction of the feedback controller

173

Example - State Feedback Control Design

3.79 11.36 0 &= x x + 0.26 u 0.2 0.2

y = [0.75 0.75]x
For the circuit, let

}
}

Physics Based SSF for circuit

u
2

Kx

State Feedback law

Find K so that the closed loop poles are at s = -5

d ( s ) ( s d ( A) ( A

5)

}
2

n 2 A
2

and desired closed loop poles at s = -5

5 I)

10 .A n 2
1

25 .I

Find

d ( A)

( A , B ) ( B AB ) K (0 1 ) . 0 0.26 2.85 0.052

controllability matrix

. ( A) d 2.83 r

Find K by Ackermans Formula

K ( 2.83 23.95 ) u Kx r k1 .x1

k1 k2 .x2

k2 23.95
Controller

answer

Control Ciruit Construction

OpAmp summer

   B

x2
Current Sensor

  B

R2 B
 C B x1
Voltage Sensor

R1 y(t)

r(t)
B

u(t)

Variable resistors can be adjusted to achieve the values required for

k1

and

k2

174

Time for a Movie

Click on image to start movie

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

Observer only measures u(t) and y(t)

& = 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

(an estimate for x)

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

}
}

From the Block Diagram

~ x = xx
Substitute for

}
~ y

observation error

~ & = Ax + Bu ( Ax &=x &x + LC~ x x + Bu) ~ & = ( A LC) ~ x x

Find the error dynamics

Observation error dynamics - tells us how fast

~ 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 )

Take Laplace Transform Rearrange Solve for

~ X ( s)

}
0

Definition of Inverse

Poles of the observation error system are given

0 ( s )
If the poles of quickly i.e.,

sI 0 ( s )

(A

LC)

are far in the left half plane, then quickly

~ 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.

The open-loop system will be observable as long as the matrix

CA ( A , C)

1 is full-rank i.e.,

..... ..... C

det( ( A , C ) ) 0

If the system is observable, the gain matrix "L" is calculated as follows

1 L o ( A)
1

( A , C)

0 .. 0

} Ackermans formula

o ( A) = o (s) s = A

where

o ( s )

specifies the desired location of the observer poles.

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 )

for n = 2 and s = -10.

0.8 2.4 3 ( A , C) . 0.2 1 0

}
4.321 20.19

Calculate observability matrix with n = 2

o ( A) .

Find L via Ackermans formula

check answer

~ & = ( A LC ) ~ x x
s

error dynamics tells us how fast

~ 0 x = xx
12.962 .136 s 7.038 8.72
2

sI

(A

LC)

64.571

} } }

Observer poles are determined by this formula

o ( s ) ( s o ( s ) s
2

12.962 ) .( s 20 .s

7.038 ) 10 )

Take Determinant

100 ( s

Simplify

Therefore if we build an observer then

x x

in 0.5 seconds

178

State Space Regulator

With only measurements of y(t), we can place the closed-loop poles anywhere as long as the open-loop plant H(s) is controllable and observable with the following regulator

& = Ax + L~ x y + Bu

= Cx y +r u = Kx

State-space regulator only measures r(t) and y(t)

~ 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.

State Space Regulator Block Diagram

Plant r( t)
+ -

u( t ) H( s )

y( t )

& x
1/s +

x
C

B K Controller

~ y
A L Observer

~ y = yy

179

Examining the Regulator Dynamics

& = 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

standard observer structure

& = Ax + B( r Kx ) + LC ( x x ) x = Cx y
Find the Error Dynamics by using

substitute for u(t) and for

~ y = C~ x

~ x = xx

& &=x &x ~ x

~ & = Ax + B( r Kx ) ( Ax + B( r Kx ) + LC( x x )) x ~ & = ( A LC) ~ x x

Subtract observer from closed-loop plant

simplified error dynamics

& = Ax + Br BK ( x ~ x x) & = ( A BK ) x + BK~ x x + Br

& = ( A BK ) x + BK~ x x + Br ~ & = ( A LC ) ~ x x y = Cx

Rewrite closed-loop plant in terms of

~ x

= x~ (x x)

Simplified Plant Composite Closed-loop System Dynamics

}
BK A

If

~ x =0 & = ( A BK ) x + Br x y = Cx

Define a new state variable system as follows

x z = ~ x

Acl

A 0

BK

LC

Bcl

B 0

Ccl ( C 0 )

& = Acl z + Bcl r z y = Ccl z

Y( s ) Ccl . sI R( s )

}
Acl

Compact Notation for Closed-loop System 1.

Bcl

Closed-loop TF

How many poles does this system have ?

180

Calculate the transfer function

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

Plug in Bcl and

Ccl .

Closed-loop TF looks exact like the full state feedback form

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

gives zero of closed-loop TF.

gives poles of closed-loop TF.

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)

canceled with the "n" observer zeros.

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

12 ) s 0.8 2.4 3 0.2

22 s

120

desired observer pole locations.

}
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

desired closed-loop poles

( 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

Standard Form for the State Equation

xn T

Define a new state

xn

x T .xn T.

d d x T . xn dt dt Bu

From the above definition

d x A .T .xn dt n y C .T .xn d 1 xn T .A .T .xn dt y C .T .x

n

State equation in terms of the new state

1 T .B .u

Premultiply by

d x A .x dt n n n y C .x
n n

Bn .u

Transformed State Equation

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

For some constant nonsingular matrix 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

} }

Property of the Identity Matrix

1 Hn ( s ) CT . T .( sI 1 Hn ( s ) C .T .T .( sI

A) T

.T

Rearrange

1 1 A) .T .T .B

} }

Use matrix identity

( 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

} }

Use Controllability Matrices

Use Observability Matrices

185

Below are two state space realizations for the circuit

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

3.79 11.36 0 &n = x xn + u 0.2 0.2 0.26 y = [0.75 0.75]xn

Find the matrix T which will satisfy

xn T

( A , B ) ( B AB )

0 1

1 4 0

Controllability matrix for RCF

An , Bn ( A , B) .

Bn An .Bn
1

2.954

0.26 0.052 0.339 0

Controllability matrix for Physics-based SSF

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

x1 0.339 .xn1 x2 1.286 .x

n1

3.846 .x

}
n2

State Relationships

186

Transformation to Jordan Normal Form

The Jordan Normal Form is a very special SSR because the eigenvalues and their controllability and observability characteristics can be read off of the SSR by inspection.

d x A .x dt y c .x d x A .x dt n J n yn CJ .xn

B .u

} }

State Space equation

BJ .u

Jordan Normal Form 1 1 AJ T .A .T BJ T .B

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

is a main diagonal matrix

(i.e., there must be only zeros appearing off the main diagonal)

Definition
The eigenvalues of

AJ

are always along the main diagonal

Theorem (for SISO system only)

If A is simple, then a) eigenvalue in b) eigenvalue in If A is not simple, a) eigenvalue in

AJ AJ

is controllable if the corresponding element in is observable if the corresponding element in

BJ CJ

is non zero. is non zero.

AJ AJ

is controllable if the last element of the corresponding Jordan block in

BJ CJ

is non zero. is observable if the first element of the corresponding Jordan block in

b) eigenvalue in

is non zero.

187

Rules for drawing Jordan Blocks

1) Draw vertical and horizontal lines to isolate the eigenvalues which lie on the main diagonal of 2) 3) 4)

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]

State Space equation

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

} } }

eigenvalues or poles of "A" eigenvectors of "A"

Formulate transformation matrix

Apply JNF formulas

3 0 1 &n = x x n + u 0 2 1 y = [ 2 1]xn

We can now mark the Jordan Blocks

188

Rules for drawing Jordan Blocks

1) Draw vertical and horizontal lines to isolate the eigenvalues which lie on the main diagonal of 2) 3) 4)

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

Jordan Normal Form eigenvalues are on the main diagonal

Analysis Since "A" is a simple matrix (i.e., "A" is a main diagonal matrix) a) eigenvalue in b) eigenvalue in

AJ AJ

is controllable if the corresponding element in is observable if the corresponding element in

BJ CJ

is non zero. is non zero.

Eigenvalue -3 -2

Controllable Yes Yes

Observable Yes Yes

189

Example - Find the JNF and discuss Controllability and Observability

& = Ax + Bu x y = Cx 1 1 2 1 1 1 0] A= 0 1 3 B = 3 C =[ 0 0 2 2 1 = 1 , m = 2 , 2 = 2 , m = 1 0 v1 = 1 0 1 T = 0 0
1

State Space equation

} } }

eigenvalues or poles of "A"

, 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"

Formulate transformation matrix

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

Apply JNF formulas

4 1 1 0 & n = 0 1 0 x n + 0 u x 0 0 2 1 y =[ 1 1 8]x n

We can now mark the Jordan Blocks

190

Rules for drawing Jordan Blocks

1) Draw vertical and horizontal lines to isolate the eigenvalues which lie on the main diagonal of 2) 3) 4)

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

Jordan Normal Form with eigenvalues are on the main diagonal

If "A" is not simple a) eigenvalue in

AJ AJ

is controllable if the last element of the corresponding block in

BJ CJ

is non zero. is observable if the first element of the corresponding block in

b) eigenvalue in

is non zero.

Eigenvalue 1 2

Controllable No Yes

Observable Yes Yes

191

Example - Find the JNF and discuss Controllability and Observability

& = 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]

}
} } }

State Space equation

eigenvalues or poles of "A"

eigenvectors of "A"

3 2 , v 2 = 3 , v3 = 0 0 0 2 0 0.33 0.66 0 3 0 T 1 = 0 0.33 0 0 1 0 0 1 1 1 0 1 AJ = T AT = 0 1 1 0 0 1

Formulate transformation matrix

C J = CT = [3 5 0]

0.33 1 BJ = T B = 0 0

}
}

Apply JNF formulas

0.33 1 1 0 & n = 0 1 1 x n + 0 u x 0 0 1 0 y = [3 5 0]x n

We can now mark the Jordan Blocks

192

Rules for drawing Jordan Blocks

1) Draw vertical and horizontal lines to isolate the eigenvalues which lie on the main diagonal of 2) 3) 4)

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

Jordan Normal Form with eigenvalues on the main diagonal

If "A" is not simple a) eigenvalue in

AJ AJ

is controllable if the last element of the corresponding block in

BJ CJ

is non zero. is observable if the first element of the corresponding block in

b) eigenvalue in

is non zero.

Eigenvalue 1

Controllable No

Observable Yes

193

Example - Discuss controllability and observability for the system below

0 0 0 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 0 0 0 0 1 & xn = 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 Normal Form

Rules for drawing Jordan Blocks

1) Draw vertical and horizontal lines to isolate the eigenvalues which lie on the main diagonal of 2) 3) 4)

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

is controllable if the last element of the corresponding block in

BJ CJ

is non zero. is observable if the first element of the corresponding block in

b) eigenvalue in

is non zero.

Eigenvalue -1 -1 -2 -3
A

Controllable No Yes

Observable Yes Yes

Yes Yes

Yes No

195

Polynomial Regulation
Given an open loop system

U( s )

H( s )

Y( s )

Open loop Plant

where

H( s )

B( s ) A( s )

} }

open-loop plant numerator polynomial open-loop plant denominator polynomial

Find the polynomials T(s), N(s), R(s) in the control structure

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 )

Desired closed-loop transfer function

Hd ( s )

Q( s ) P( s )

} }

desired closed-loop numerator polynomial desired closed-loop denominator polynomial

Restrictions 1) 2)

degree ( P ( s ) ) degree ( A( s ) )
Q(s) must contain the roots found in B ( s )

196

Steps for Calculating T(s), R(s) and N(s)

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)

and check restriction 2

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

Q(s) must contain the roots found in

B ( s)

Define R(s) and T(s) in terms of subpolynomials

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

Determine the degree of each polynomial

deg{T1} = deg{ A} deg{B + } 1 = L deg{T1} = L

By definition

deg{R1} = deg{T1} = L deg{N } = deg{ A} 1 = C } By definition deg{N } = C

197

Step 5

Define the structure of

T1 ( s ) , R1 ( s ) , N( s )

T1 ( s ) ( s R1 ( s ) s
L c

a)

}
1

"a" is far in the left-half plane and "L" is known 1 1

rL

.s L
1

...... ......

r0 n0 ri , ni ri , ni

} }

"L" is known "C" is known are unknown

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

Find T(s), R(s) and N(s) such that

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

Restrictions 1) Step 2 Find

degree ( P ( s ) ) degree ( A( s ) ) 2

OK

B + ( s)

and

B ( s)

and check restriction 2

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) must contain the roots found in B ( s )

Q( s ) ( s
Step 3

1)

B ( s ) = ( s 1)

OK

Define R(s) and T(s) in terms of subpolynomials

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 )

are unknown at this point.

199

Step 4

Determine the degree of each polynomial

deg{T1 } = deg{ A} deg{B + } 1 = 2 0 1 = 1 deg{T1 } = 1 deg{R1} = deg{T1} = 1 deg{N } = deg{ A} 1 = 2 1 = 1 deg{N } = 1

Step 5 Poles of Let a = 10 Define the structure of

L 1

C 1

T1 ( s ) , R1 ( s ) , N( s )

T1

should be 10 times bigger than the largest pole of P(s)

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

Solve the Diophantine Equation to determine

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

} }

Solve for coefficients Gives

( 31 s

10 )

R1 ( s )

and

N( s )

Construct R(s) and T(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

Block Diagram for closed-loop system

V( s )

s s

10 43

(s s
3

43 ) .( s
2

1) 10

Y( s )

12 s

21 s

Feedforward Feedback Combination

V( s ) s
3

(s

10 ) .( s
2

1) 10

Y( s )

12 s

21 s

} } }

Combine cascade blocks

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 )

Cancel observer pole

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

is usually arrived at by examining the "physics" of the problem

or by setting f(x,u) = 0 and solving for x. Note:

xe

is an n x 1 vector; so youneed "n" values for the equlibrium point

202

Example

m L g

One link robot

&& mgL sin u = mL

u 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

Step 1 - Write the dynamics in the

form.

&& mgL sin u = mL

original dynamics

&& = g sin( ) + 1 u } rearrange mL && x1 = From &2 = x x2 = & problem } &1 = x2 x statement

&1 = x2 x 1 & 2 = g sin( x1 ) + x u mL

use above information to rewrite dynamics

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

}
}

use formula with n = 2

x2 f & = f ( x, u ) = 1 = x f 2 g (sin( x1 )) +
f x
0 g .cos x1 1 0

from above work

1 mL

}
}

substitute for

f1 and f2

into the formula

f = 0 u

1 ml

substitute for

f1 and f2

into the formula

Step 3) Substitute the equilibrium point to find A and B

x1e e 0 x2e

& =

=0
T

x xe = 1e = &e x2 e e 0 1 = + g 0
0 = 1 mL

Equilibrium Point from problem statement

f A= x
f B= u

x = xe

Plug-in equilibrium point and transpose

x = xe

Step 4)

Find "C" from problem statement

y = y = x1
y (1 0 ) . C (1 0 ) x1 x2

Since

x1

by definition

204

Linearized System for One Link Robot

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

Open Loop TF formula

}
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

Laplace Transform } Inverse Impulse Response

}
t

Unstable Impulse Response

205

Control Design

0 1 0 &= + x x 1 / mLu g 0 y = [1 0]x

Find K in

}
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 )

Formula for closed-loop denominator

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

Note the control is given by

Kx

where

x = &

so

&+r u = mL( g + 1) 2mL

control which must be implemented

206

Computer Implementation Plant

r( t)

+ -

u(t)

Encoder

y( t )

mL( g
+

1)

&
2 mL

Numerical Differentiator

Computer

&+r u = mL( g + 1) 2mL

Control

207

Same Example

m L g

What if

&

is not available for measurements ?

Solution:

Use a polynomial regulator

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

Find T(s), R(s) and N(s) such that

Y( s ) 1 V( s ) ( s 1 ) 2

Objective: Place both closed-loop poles at s= -1

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)

and check restriction 2

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) must contain the roots found in B ( s )

Q( s ) 1
Step 3

B ( s) = 1

OK

Define R(s) and T(s) in terms of subpolynomials

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

are unknown at this point.

Determine the degree of each polynomial

deg{T1 } = deg{ A} deg{B + } 1 = 2 0 1 = 1 deg{T1 } = 1 deg{R1} = deg{T1} = 1 deg{N } = deg{ A} 1 = 2 1 = 1 deg{N } = 1

Step 5 Poles of Let a = 10 Define the structure of

L 1

C 1

T1 ( s ) , R1 ( s ) , N( s )

T1

should be 10 times bigger than the largest pole of P(s)

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

are unknown at this point

209

Step 6

Solve the Diophantine Equation to determine

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

Solve for coefficients

R1 ( s ) s N( s ) ( g
Step 7

12 21 ) s ( 12 g 10 )

}
s + 12 mL

Gives

R1 ( s )

and

N( s )

Construct R(s) and T(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

Check your answer

V( s )

(s s

10 ) 12 mL

(s (s 12 ) . s
2

12 ) . ( 21

1 mL g) s ( 12 g 10 )

Y( s )

.... Feedforward Feedback Combination

V( s ) s
3

s 12 s
2

10 21 s 10

Y( s )

Combine cascade blocks

V( s ) (s

(s 10 ) . s

10 )
2

Y( s ) 1

2s

} }

Factor denominator

V( s ) (s

1 1)
2

Y( s )

Cancel observer pole

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.

The skeleton-in-the-closet of L Transforms + 1 L [1(t)] = s = L [1(t)] and L [ x ] = s( L x) + + +

we obtain

- x(0+)

L+ [ (t )] = s.s 1 1 = 0 = L+ [0]

This may be a surprise, though it is consistent with the direct calculation

(t )] L+ [
L-

0+

(t )e

st

dt

=0

The expected formula for the Laplace transforms of from the use of the formulas. =

(t )

is 1. But this actually follows

which also agrees with the direct evaluation. More detailed discussions of the

(t )] L -[

=L

[d1(t ) / dt ]

s.s 1 0 = 1

transform can be found in many places, e.g., L.A. Zadeh and

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

1(t ) 1(t ) d = 1(t ) 0 dt

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

A pictorial interpretation using sequences is provided by the figure below. As

, 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