Classical Methods of Control Systems (Shiakolas)

BLOCK DIAGRAM REPRESENTATION
OPEN LOOP
PLANT
INPUT
x(t )
X (s)
G (s )
OUTPUT
y (t )
Y ( s)
Gc
R
Y ( s) = G ( s) X ( s)
Y (s)
G ( s) =
X (s)
CONTROLLER
GP
PLANT
Error, E = R Y = 0 is desired
X = RGC
Y = XGP
Y = RGC GP
Now , E = R RGC GP = 0
R(1 GC GP ) = 0
Q R 0, 1 GC GP = 0
1
GC =
GP
CLOSED LOOP
CONTROLLER
PLANT
G1
MEASUREMENT
Y = UG1
U = EC
E = R YH
Y
CG1
=
R 1+ CG1H
Denominator : 1 + CG1 H = 1 + GH
{
G
1 + GH Characteristic polynomial
1 + GH = 0 Characteristic equation (CE)
Y
G
=
R 1 + GH
If CE is known, we can determine,
order of system
eigen values
stability
response characteristics
EXAMPLE
Applying Conservation of Angular
Momentum
h& = mgl sin
h = ml &
h& = ml 2&&
h = I&
ml 2&& + mgl sin =

1
mgl
&
&
+ 2 sin = 2
ml
ml
ml 2&& = mgl sin

Need to linearize
Operating point
Static equilibrium point
= (t )
g
1
&
&
+ sin = 2
l
ml
= (t )
& = && = 0
Static Equilibrium,
0 , 0
1
sin 0 =
0
0
2
gml
gml
Assume 0 = 0
180o
sin 0 = 0
0 = 0 or 180
2005 P. S. Shiakolas
0o
ME 5303 Classical Methods of Control

Systems Analysis & Synthesis
LINEARIZATION Taylors Series Expansion
g
1
&
&
+ sin 2 = 0 = f &&, g , l , m, ,
l
ml
f && &&
f
f
(
(
0 +
l0 )
f f0 =
g g0 ) +
l
424
3 l 123
&& 0
g 0 1
0
g =0
l =0
f
f
f
(1
(
(
+
m m0 ) +
0 ) +
0 )
3 m
m 0 424
0
0
m = 0
1
g
&
&
f = 0 = 1 +
+ cos
l
0
ml 2 0

1
g
&
&
+ cos =

0
l
ml 2 0
If the equilibrium point is = 0o
1
g
&
&
+ cos 0 =
ml 2
g
1
&
&
+ =
l
ml 2

If the equilibrium point is = 180o

1
g
&
&
+ cos180 =
l
ml 2
1
g
&
&
=
l
ml 2

LAPLACE TRANSFORM (Assume Zero ICs)

g
1
&
&
+ =
Laplace Transform
2
l
ml
g
1
( s ) s + ( s ) =
l
ml 2
2
1
2 g
s + = 2
l
ml
1 1
=
ml 2 s 2 + g
PLANT
or
1 1
=
ml 2 s 2 + g
l

g
s + = 0 characteristic equation
l
g
Im
eigen values s =
i
l
2
C.E. for 0 = 180o
Unstable region
g
2 g
s =0 s=
l
l
If the eigen values have
positive real parts, then the system is unstable
negative real parts, then the system is stable
zero real part, then the system is marginally stable

Re

g
1
&
&
+ =
l
ml 2
State space
representation
Set of first order DFQs

z1 =
z 2 = z&1 = & z&1 = z 2
Number of DFQs = Order of system
g
1
z&2 + z1 =
2
l
ml
g
1
z&2 = z1 +
2
l
ml

z&1 = 0 z1 + 1z2 + 0
g
1
z&2 = z1 + 0 z2 +
2
l
ml
x& = Ax + Bu where
z&1 0
= g
z&
2 l
Output :
A = coefficient matrix square

B = order of the system
1 z 0
1
+ 1
0 z 2
2 ml
y = Cx + Du
C.E. det (SI A) = 0

s
det
0
s
det g
l
0
g
s
l
1
=0
0
1
s
=0
g
s + =0
l
2

PROJECT
PENDULUM
mc
0 = 180o
mr

MAGLEV
electro
magnet
sensor
object mass, m
L
Magnet
Fmagnet
mass
Fmagnet = f (k , I , x )
weight

Derive system dynamic equations

Linearize, if needed
Derive state space representation
Find transfer function
Find characteristic equation

LAPLACE TRANSFORMS
Final Value Theorem (FVT)
f () lim s F ( s )
s 0
Condition: All the poles of F(s) other than s = 0 must
have negative non - zero real parts.
Poles: Are points at which the f(t) or its derivatives
infinity
Zeros: Are points at which the f(t) equals zero

(
s + 2 )(s + 10 )
G ( s) =
s(s + 1)(s + 7 )(s + 17 )2
2 zeros, 5 poles
Im
Re
-17
-10
-7
-2 -1 0

TRANSFER FUNCTIONS (T.F.)

L[output ]
T.F. = G ( s ) =
L[input ] Zero I.C.'s
Y ( s ) bm s m + bm 1s m 1 + ... + b1s + b0
=
, nm
X (s)
s n + an 1s n 1 + ... + a1s + a0
n Order of the system
Applicable only for Linear Time Invariant
(LTI) Systems
Independent of input (magnitude and time)

BLOCK DIAGRAMS
Output
Input
a1
a2
Description of component
or subsystem
Branching point
b1
Signal
a3
Summing JCT: a1 + a2 a3 = b1

D
E
G
H
PRINCIPLE OF SUPERPOSITION (LTI)
C = C R, D = 0 +
R
CR
G
H
D
P
G
H
GP
CR =
R
1 + GPH
C D, R = 0
P
CD =
D
1 + PHG

CD
GP
P
C=
R+
D
1 + GPH
1 + GPH
1 + GPH Characteristic polynomial
STATE SPACE REPRESENTATION

System of linear DFQs System of First Order
linear DFQs
k
m
f (t )
b

m&x& + bx& + kx = f (t )
2 nd order linear DFQ 2 first order linear

DFQ in state space
State space variables : z1, z2
z1 = x
z2 = z&1 = x&
z&2 = &z&1 = &x&
m z&2 + bz2 + kz1 = f
1
z&2 = ( kz1 bz2 + f )
m
& z&1 = z2

x& = Ax + Bu
y = Cx + Du
Y
X
( x, y ) 2 variables
l 2 = x 2 + y 2 Constraint Equation
Only one variable for system state

x& = A x + Bu
L sIX x0 = AX BU
(sI A)X = BU + x0
X = (sI A)1{BU + x0 }

y = C x + Du
L Y = CX + DU
Y = C (sI A)1{BU + x0 } + DU
Y = C (sI A)1 B + D U + C (sI A)1 x0
Denominator of the above eq. sI A

Therefore, sI A is the Characteristic Polynomial
& sI A = 0 is the Characteristic Equation of the
system.

Time Step for Numerical Integration

Dynamics Differential Equations
Linearize
State Space
Block Diagram State Space form

INTEGRATOR
x&
x1
x2
1
x1 = x2
s
x1 = x&2
State variables at output of integrators
1
f ( s)

Example:
U
10
s ( s + 2)
1
s +1
1
s
X3
X2
10
s ( s + 2)
1
s +1
X1
Y
FF
CE :
=
U 1 + FF FB
10
1
Denominator : 1 +
=0
s ( s + 2) s + 1
(s 2 + 2s )(s + 1) + 10 = 0 3rd Order
10
X2
= X 1 10 x2 = x&1 + 2 x1
s+2
1
X1
x1 = x&3 + x3
= X3
s +1
1
(U X 3 ) = X 2 u x3 = x&2
s

x&1 = 2 x1 + 10 x2 + 0 x3 + 0u
x&2 = 0 x1 + 0 x2 1x3 + 1u
x&3 = 1x1 + 0 x2 1x3 + 0u
x&1 2

x& = 0
2
x& 1
3
10
0
0
0 x1 0

1 x2 + 1 u

1 x3 0
Z = 1x1 + 0 x2 + 0 x3 + 0u
Y = [1
0] x + 0u
Transfer Function has zeros or derivatives of input
Y
f1 ( s )
=
U f 2 (s)
X
s+a
=
U s 2 + 2n s + n2
&x& + 2n x& + n2 x = u& + au
Converting into state space
z1 = x z2 = x& = z&1

z&1 = z2
z&2 = n2 z1 2n z2 + au + u&
TF
X X Z
TF = =
U ZU
1
DEN
NUM

State space representation w.r.t. variable z
X
s+a
=
U s 2 + 2n s + n2
From the block diagram,
Z
X
1
=
= s+a
&
2
2
U s + 2n s + n
Z
Z
&z& + 2n z& + n2 z = u
U
Converting into state space,
=z
= & = z&
1

&1 = 01 + 12 + 0u
&2 = n21 2n2 1u
&1 0
=
& 2
2 n
1 0
+ u
2n 2 1
1
X
x = z& + az
Z
x = a1 + 12 + 0u

FEEDBACK CONTROL SYSTEM

CHARACTERISTICS
STATE
SPACE
BLOCK
DIAGRAM
TRANSFER
FUNCTION

PLANT
CONTROLLER
G1
MEASUREMENT
F = EG
X
GP
=
R 1 + GPH
CE :1 + GPH = 0
Find such that

the block diagrams
are equivalent.
Unity Feedback

EFFECTS OF FEEDBACK
Affects
closed loop behavior
stability
bandwidth
overall gain
Reduces sensitivity of the output to
disturbances and system/plant parameter
changes.
Improve the steady state error
Ease control of transient response

GENERAL PROPERTIES OF F.B.C.S.

T
R
G
H
GP
P
X=
R+
T
1 + GPH
1 + GPH
CE :1 + GPH = 0
E = Reference Signal Output Signal
E = R XH

DEP
SENSITIVITY
L2
L1
Output , C or X
C = f (G, P, H , T , R )
Linearize
C
C
C
G +
P +
H
C =
H 0
G 0
P 0
C
C
+
R +
T
R 0
T 0

IND
OPEN LOOP
H =0
CLOSED LOOP
GP
P
R+
T
C=
1 + GPH
1 + GPH
C = GPR + PT
SENSITIVITY TO DISTURBANCES
R = 0 ; T 0
G = P = H = 0
P
C
C =
T
C =
T
1 + GPH
T
0
C = PT
P
C
T
1
+
GPH
if R = 0,
=
P
C
T
1 + GPH
SENSITIVITY TO SENSOR GAIN
GP
C=
R
1 + GPH
GPGPR
C
C =
H
2
H
(1 + GPH )
C
Normalize
C
C G 2 P 2 R (1 + GPH )2
=
H
C
GP 1 + GPH R

C
GP
=
H
C 1 + GPH
if 1 + GPH >> 0
C
H
GP
H
C
GPH
H
C
H
C H

Abs Value
=
C
H
C
H
One to one mapping

EXAMPLE
G=
R unit step
n2
s(s + 2 n )
= 0.6 & n = 5 rad/sec

Find the time domain performance specifications.

Rise time t r =
d
t r = 0.554 sec
d = n 1 2
d = 4 rad/sec
= atan
= 0.927 rad


Peak time t r =
= 0.785 sec
d
Settling time 2% t s =
5% t s =
n
3
Max overshoot M p = e
= 1.333 sec
= 1.000 sec
= 0.095

EXAMPLE
k
G=
s(s + 1)
H = 1 + kh s
Determine k & k h s.t.
1
s (s + 1)
max. overshoot = 0.2 units

peak time = 1.0 sec
R unit step
Obtain tr , t s

Mp =e
= 0.456
tp =
d = 3.14
d
d = n 1 2 n = 3.53
CE : 1 + GH = 0
k
(1 + kh s ) = 0
1+
s (s + 1)
s 2 + (1 + kkh )s + k = 0

Compare with standard 2 nd order

s 2 + 2 n s + n2 = 0
n2 = k k = 12.5
2 n 1
2 n = 1 + kkh k h =
= 0.178
k

tr =
= 0.65 sec
d
t s = (2% )
= 2.48 sec ; (5% )
= 1.86 sec

EXAMPLE
Negative Feedback C.L.
k (s + 2 )
GH =
s (s 2 )
a) k such that of C.L. poles = 0.707

b) k such that C.L. poles are on Im. axis
1 + GH = 0 s 2 + (k 2 )s + 2k = 0
n2 = 2k n = 2k
2 n = k 2 2 2k = k 2
a)
= 0.707 2(0.707 ) 2k = k 2

k 2 8k + 4 = 0
Im
Re
b) C.L. poles on Im axis
Re part of eigen value = 0

s1,2 = j (s s1 )(s s2 ) = 0
s 2 + 2 = 0 2 n = 0 = k 2
k = 2

Im
constant
line
k =2
k = 0.53
j d
Re
1 2
d
= atan n
= atan

n
n
1 2
= atan

STANDARD CONTROLLERS
R
GC
GP
H
ONE TERM CONTROLLER
Gc = k
H =kf

TWO TERM CONTROLLERS

a) PI: Proportional Integral
1 kc i s + 1
=
Gc = kc 1 +
is i is
b) PD: Proportional Derivative
Gc = k p + s = kc ( d s + 1)
THREE TERM CONTROLLERS

PID: Proportional Integral Derivative
Gc = kc 1 + d s +
is

GC
GP
H
1
Gp =
s (s + 1)
No Controller
1
C
=
R s2 + 1 s + 1

PD Controller
PD Gc = kc (1 + d s )
kc (1 + d s )
GcG p
C
=
=
R 1 + GcG p
2 1 + kc d
s +

PI Controller
kc
s +
PI Gc = kc 1 +
is
kc (1 + i s )
(
i )
C
=
R s 3 + 1 s 2 + kc s + kc

Two Term Velocity Feedback
H = 1+ k f s
kc
KC
K cG p
C
=
=
R 1 + K cG p
2 1 + kc k f
s +

R
kc
s + 1
GP
H
kc
s +
GP
kf
H

STEADY STATE ERROR TABLE

Type
0
1
Step
A
1+ k p
0
INPUT
Ramp Parabolic
A
kv
A
ka

EXAMPLE
Design a controller that would satisfy :
ess for ramp imput < 10% of input

of dominant poles 0.707
settling time (2% ) 3 sec
k1
G=
s (s + 2 )
H = 1 + k2 s
R=
A
s2
G
H
C
k1
=
R s 2 + (2 + k1k 2 )s + k1
Type of system = Type 1

GH =
k1
s1 (s + 2 )
(1 + k2 s )
Assume FVT applicable

ess = lim sE ( s ) = lim s(1 TF )R
s 0
s 0
+
+
s
k
k
2
1 2
= lim A
s 0 s 2 + (2 + k1k 2 )s + k1
A(2 + k1k 2 ) 10
ess =
<
A
k1
100

CE : s + (2 + k1k 2 )s + k1 = 0
2
n2 = k1 & 2 n = 2 + k1k 2
n = k1
2 n
2 + k1k 2
< 0.1
< 0.1
k1
k1
2(0.707)n
Let = 0.707
< 0.1
k1

2
n > 14.14 k1 = n = 199.94 200
2 n = 2 + k1k 2 k 2 = 0.09
Check third requirement
4
t s (2% ) =
=
= 0.4 < 3 sec
n (0.707)(14.14)
4

STABILITY
STABLE
t
Im
NEUTRAL
Re
UNSTABLE
t

RUTH-HURWITZ CRITERION
CE : an s n + an 1s n 1 + an 2 s n 2 + ... + a1s + a0 = 0
TESTS FOR APPLYING R.H.
a0 0 ; All ai Real
If any of the coefficients are zero or change sign,
then system is unstable

ROUTH TABLE
Power
Coefficients
sn
an
an 2
an 4
s n 1
an 1 an 3
an 5
sn2
b1
b2
s n 3
c1
c2
b3
M
s0

1 an
b1 =
an 1 an 1
an 2
1 an
b2 =
an 1 an 1
an 4
an 3
an 5
1 an 1
c1 =
b1 b1
an 3
1 an 1
c2 =
b1 b1
an 5
b2
b3
RH States The number of poles of CE with

positive real parts (unstable poles) is equal to the
number of sign changes in the first column of the
Routh array.
Case 1 2 nd Order System : a2 s 2 + a1s + a0 = 0
a2
a0
a1
s0
b1 = a0
1 a2
b1 =
a1 a1
a0
0
1
( a1a0 )
b1 =
a1
b1 = a0

CE : a3 s + a2 s + a1s + a0 = 0
a3
a1
3
1
s
a3
a1
b1 =
a
a0
2 a2
2
s
a2 a0
1
(a0a3 a1a2 )
b1 =
s1
b1
b2
a2
s0
c1 = a0
1 a2
c1 =
b1 b1
a0
0
1
(a1a2 a0a3 )
=
a2
= a0
For stable system,

b1 > 0 a1a2 > a0 a3

CE : s 3 + s 2 + 2 s + 24 = 0
2
1 1
3
s
1 2
b1 =
= 22
1 1 24
s2
1 24
24
1 1
= 24
s1
b1 = 22 c1 = b
0
1 b1
s0
c1 = 24
System is unstable as there is a

sign change in the first column
# of unstable poles = # of sign changes in

the first column
Eigen values : s1,2 = +1 j 7 & s3 = 3


CL feedback system with negative feedback
Gp =
1
a3s 3 + a2 s 2 + a1s + a0
Gc = kc
H =1
Will this system be stable or unstable?

Is there a limiting value for kc for stability?
CE : 1 + Gc G p = 0
a3s 3 + a2 s 2 + a1s + a0 + kc = 0

Applicable ? All coefficients > 0
a0 + kc > 0 kc > a0
a1a2 a3a0 a3kc
b1 =
>0
a2
c1 = a0 + kc kc > a0
a1a2 a3a0
a3kc a3a0 a1a2
>
kc <
a3
a2
a2
a1a2 a3a0
a0 < kc <
a3

CE : s 4 + ks 3 + s 2 + s + 1 = 0
Determine the range of k for stability
4
s3
s2
b1
s1
c1
s0
d1
b2
1 1
b1 =
k k
1 1
b2 =
k k
1
= (k 1)
1 k
=1
k2
c1 = 1
k 1
d1 = 1

1
b1 = (k 1) > 0 k > 1
k
k2
k 1 k 2
c1 = 1
>0
>0
k 1
k 1
negative
1 + k (1 k )
>0
k 1
positive
The above range of inequality is false and
therefore there is no range of k that yields us
a stable system.
Hence, system is unstable.

Case 2 Zero in first column while all other

elements are non-zero
5
CE: s + 2 s + 2 s + 4 s + 11s + 10 = 0
s5
11
s4
10
s2
1
s
s
b1
c1
d1
e1
b2
12
1 1
b1 =
2 2
1 1
b2 =
2 2
11
1 1
c1 =
b1 b1
=0
10
11
b2
=6
Case 3 Whole Row full of zeros

3
CE: s + s + 2 s + 2 = 0
s1 row is all zeros
s3
s2
s1
b1 = 0
s0
c1
Form the auxillary polynomial
1s 2 + 2 s 0 s 2 + 2
Eigen values of s 2 + 2 is 2 j
which are symmetric about
the real axis.
Differentiate auxillary polynomial 2 s


CE : s 4 + s 3 + 3s 2 + 2 s + 2 = 0
s4
s
s2
b1
b2
s1
c1 = 2
s0
d1
b1 = 1
b2 = 2
c1 = 0
Q = b1s 2 + b2 = s 2 + 2
dQ
= 2s
ds
1 b1
d1 =
c1 c1
b2
0
= b2

CE: s + 2 s + 24 s + 48s 25s 50 = 0

s5
24
25
b1 = b2 = 0
50
Q = 2 s 4 + 48s 2 50
48
s3
b1
b2
s2
c1
b1 = 8 & b2 = 96
s1
d1
c1 = 24 & c2 = 50
s0
e1
d1 = 112.7 & e1 = 50
dQ
= 8s 3 + 96 s
ds
System is unstable

CE : s 3 + 2 s 2 + 4 s + k = 0
s3
s2
s
s
b1
c1
1 1
b1 =
2 2
8k
=
>0
2
k
1 2
c1 =
b1 b1
=k >0
8k > 0 k < 8
0<k <8

Gp =
Gp
k
s ( s 2 + s + 1)( s + 2)
Find k for stability
Apply RH for CE : s 4 + 4 = 0

Gc
Gp
Gc = k
Steady State Error
Gp =
Type = 0; Step Input (unity)
1
ess =
1+ k p
2
s 3 + 4 s 2 + 5s + 2
k p = lim GcG p = k
s 0
1
ess =
1+ k
1
1
ess < 2% input
<
k > 49
1 + k 50

CE: s 3 + 4 s 2 + 5s + 2 + 2k = 0
2 + 2 k > 0 k > 1
18 2k
3
b1 =
>0
s
1 5
4
c1 = 2 + 2k
s2
s1
b1
s0
18 2k
>0 k <9
4
c1
For stability, 1 < k < 9 which
2 + 2k
contradicts k value from steady state error.


p - controller cannot be used.

Investigate PI controller.
Gc =
k p s + kI
s
CE : s 4 + 4 s 3 + 5s 2 + 2 + 2k p s + 2k I = 0
2 + 2k p > 0 k p > 1
kI > 0

s4
s3
4 2 + 2k p
s2
b1
s1
c1
s0
d1
b2
2k I
b1 =
18 2k p
4
>0
kp < 9
b2 = 2k I
d1 = 2k I > 0
[(
)(
4
c1 =
1 + k p 9 k p 8k I
18 2k p

ROOT LOCUS (RL)

Graphical technique that will allow us to plot
the path or locus of the eigen values as one
parameter changes (0 to )
RL has the ability to
Determine closed loop system behavior
from open loop conditions
Determine the effects of
one parameter
qualitatively

Complex Algebra Review

ROOT LOCUS (RL)

Graphical technique that will allow us to plot
the path or locus of the eigen values as one
parameter changes (0 to )
RL has the ability to
Determine closed loop system behavior
from open loop conditions
Determine the effects of
one parameter
qualitatively

Complex Algebra Review
A = + j = A e j
Im

= atan2 = A

A = 2 + 2
A1 e j1
A1
A1 j (1 2 )
=
=
e
j
A2 A2 e 2 A2
A1 A2 = (1 + 1 j ) ( 2 + 2 j )
= (1 2 ) + (1 2 ) j

Re
Background on RL
CLTF
C
G
=
R 1 + GH
CE : 1 + GH = 0
G
H
GH: Ratio of polynomials in s

Where s: complex quantity
s = j
1 + GH s = j = 0
GH s = j = 1

Magnitude GH s = j = 1
Angle
GH s = j = ( 1 + 0 j )
= 180(2n + 1) n = 0,1, 2...
EXAMPLE
Im
k
GH =
s +
GH =
( + ) + j
s = j

Re
GH =
( + )2 + 2
GH = = atan
= 180(2n + 1)
k
+1 = 0
GH + 1 = 0
s +
s + (k + ) = 0 s = (k + )
k =
x
Im
k =0
x

Re
Properties of RL
GH = 1 where G = k g
kg
Ng
Dg
kf
Nf
Df
Ng
Dg
& H =kf
= 1 k = k g k f =
Nf
Df
Ng N f
Dg D f
Dg D f = 0
if k = 0
Open loop poles
N g N f =
Dg D f =
if k =
Open loop zeros
N g N f =

OLTF :
CLTF :
Ng N f
Dg D f
C
=
R
kg
Ng
kg
Ng
Dg
Dg
=
Dg D f + kN g N f
Ng N f
1+ k
Dg D f
Dg D f
k 0 : CL poles approaches OL poles

k : CL poles approaches OL zeros

EXAMPLE
k
GH =
s +1
0k
CE : s + (1 + k ) = 0
OL poles : s = 1
OL zeros :
k =
x
Im
k =0
x
-1
Re
EXAMPLE
k (0.4 s + 1) OL pole : s = 1
GH =
s +1
OL zero : s = 2.5
Im
TP2
TP3
TP1
TP5
-1
Re
TP4
GH = k + 0.4 s + 1 s + 1
GH = 0 + = 180(2n + 1)

CE : (1 + 0.4k )s + 1 + k = 0
1+ k
s=
1 + 0.4k
k = 0 s = 1
k = s = 2.5
k = 5 s = 2
101
k = 100 s =
41
1001
k =0 s=
401

GH s = 1.5 = 1
k (0.4 s + 1)
=1
s + 1 s = 1.5
0.8k = 1
k = 1.25

ROOT LOCUS (Cont)

Starts at O.L. poles k varies from 0 to
Ends at O.L. zeros.
1 + kGH = 0 where kGH is OLTF

k
OLTF: GH =
s (s + 1)
OL Poles :
s = 1
OL Zeros : none
CE : s 2 + s + k = 0
1 1
1 1 4k
1 4k
s1,2 =
=
2
2 2
s=0

-1
-1/2
Im
1
0

Re
CL poles
1
4
1
2
1 1
+ j
2 2
10
1
39
+
j
2
2
1000
1
3999
+
j
2
2
1 1
j
2 2
1
39

j
2
2
1
3999

j
2
2

Identify controller gain k to yield eigen value at

1
s = 3j
2
GH for all s = 1 = 1
k
=1
s (s + 1) s = 1 + 3 j
2
find k .

1
0 < k < , the eigen values are on the the Re - axis
4
i.e. they behave as the first order system.
(s + )(s + ),
> 1.
1
k = , the eigen values are on the Re - axis
4
& repeating. (s + )2 , = 1.
1
k > , the eigen values are complex conjugates.
4
< 1.

Im
s 2 + 2 n s + n2 = 0
-1
-1/2
d n 1 2 1 2
tan =
=
=
n
n


Re
EXAMPLE
= 0.6 (desired damping)
1 0.6 2
tan =
= 1.333 = 53.1o
0.6
1
n = n = 0.833
2
d = n 1 2 = 0.6664
Point of interest Desired eigen value
1
Find k to yield s = 0.6664 j
2

RULES FOR DRAWING ROOT LOCUS

Example
k
OLTF : GH =
(s + 1)(s + 2)(s + 3)
Find OL poles : 3 (1, 2, 3)

OL zeros : 0
Draw OL poles & zeros
Start OL poles (k = 0 )
End OL zeros (k )
# Loci = 3

Symmetry about Re axis

n # OL poles = 3
m # OL zeros = 0
# asymptotes = na = n m = 3
(loci going to )
Angle of asymptotes with the Re axis
180(2q + 1)
q =
nm
180(2q + 1)
=
= 60(2q + 1)
3

60
1 180
2 300
Intersection of asymptotes with Re axis
(Centroid of
poles & zeros )
poles zeros
a =
nm
(
- 1 - 2 - 3) 0
=
= 2
3

Loci on Re axis : Loci lie on Re - axis in

regions where there is an odd number of
poles and zeros (on Re - axis) to the right of it.
Intersection of RL with Im - axis will yield
limits of gain for stability
Routh - Hurwitz
On Im - axis s = jw
Substitute s = jw in CE
CE : s 3 + 6 s 2 + 11s + 6 + k = 0

( jw)3 + 6( jw)2 + 11( jw) + 6 + k = 0
(
)
Im ( 3 + 11 ) j = 0
(11 2 ) = 0 = 0
Re 6 2 + 6 + k = 0
= 11
( )
6 11 + 6 + k = 0 k = 60
Breakin/Breakaway point
CE :1 + kGH = 0 = f ( s )

A + kB = 0 = f ( s )
( A = A( s), B = B( s) )
A
k =
B
dk
d A
min/ max
= =0
ds
ds B
The value of s is breakin/away pts if k
for that s is positive.

RULES FOR DRAWING ROOT LOCUS (Cont..)
A
k=
B
A = (s + 1)(s + 2)(s + 3)
B =1
dk
d s 3 + 6 s 2 + 11s + 6
=0
=
1
ds
ds
= 3s 2 + 12 s + 11 = 0
s1 = 2.6 & s2 = 1.4
1. Evaluate k breakin/away. Reject if k < 0
2. Examine the Re axis

ROOT
LOCUS
Root Locus
5
4
Constant
line
IMAGINARY
AXIS
Imaginary Axis
= 11, k = 60
1
0
-1
s = 1.4
k = 0.384
-2
-3
-4
-5
-8
-7
-6
-5
-4
-3
-2
-1
Real Axis
REAL AXIS

Example GH =
10(s + )
2
s + 5s + 6
0< <
N (s)
k
= GH
D( s)
CE : 1 + GH = 0
s 2 + 5s + 6 + 10 s + 10 = 0
(s 2 + 15s + 6)+ 10 = 0
1+
10
s 2 + 15s + 6
=0

n=2 ; m=0
na = n m = 2
180(2q + 1)
= 90(2q + 1) 90, 270
a =
nm
PZ
a =
= 4.5
nm
Breakin/away point
s 2 + 15s + 6
d
=
= (2 s + 15)
ds
10
s = 7.5

Root Locus
Imaginary Axis
4
2
0
-2
-4
-6
-15
-10
-5
Real Axis

Example
GH =
n=3
m=3
k (s + 3)(s + 1 3 j )
s (s + 1)(s + 2 )3 (s + 4 )(s + 5 2 j )
(0,1, 2,2,2,4,5 2 j )
( 3,1 3 j )
na = n m = 5
180(2q + 1)
a =
= 36(2q + 1) 36, 108, 180
nm
PZ
= 3.2
a =
nm

Root Locus
15
Imaginary Axis
10
5
0
-5
-10
-15
-20
-15
-10
-5
Real Axis

10
Example
Gp =
s 3
H =1
s 2 + 5s + 6
Gc = 1 (proportional)
Is the system stable or unstable?
OLTF :
k (s 3)
2
s + 5s + 6
assuming Gc = k
How to obtain a quadratic type response?

Root Locus
1
Imaginary Axis
0.5
-0.5
-1
-14
-12
-10
-8
-6
-4
Real Axis
-2

STANDARD CONTROLLERS
P: k
No OLP/OLZ
is +1
PI : k
is
1 OLZ
1 OLP (origin)
PD : k ( d s + 1)
1 OLZ
d i s 2 + i s + 1
PID : k
s
i
2 OLZ
1 OLP (origin)

Im
PI
x
Re
Im
0 < k < klimit
Im
Re

Re
PD
Im
Im
constant
x
x
Re
Re
Im
Im
x
x
Re
Re
Re
Im
0 < k < klimit


PID
Im
Im
Re
Marginally Stable unstable

Re
Example
(
s + 2 )(s + 7 )
GH = kc
(s 1)(s + 1)(s + 10)(s + 3)
Root Locus
m=2
na = n m = 2
15
180(2q + 1)
a =
nm
PZ
a =
nm
20
Imaginary axis
n=4
Root Locus
25
10
5
0
-5
-10
-15
-20
-25
-10
-8
-6
-4
Real axis

-2
Example
R
1/s
A
1 + Ak
20
A=
(s + 1)(s + 4)
Draw Root Locus

Determine the value of k such that the damping ratio
of the dominant pole C.L. poles = 0.4

A 1
C
CLTF :
= 1 + Ak s
R 1+ A 1
1 + Ak s
CE :
A
1+
1 + Ak
1
=0
s
(s3 + 5s 2 + 4s + 20)+ 20ks = 0

1+ K
s
3
s + 5s + 4 s + 20
=0

OLTF for RL :
s
K
(s + 2 j )(s 2 j )(s + 5)
Root Locus
15
n=3
m =1
na = n m = 2
a = 90(2q + 1)
a = 2.5
Imaginary Axis
10
5
0
-5
-10
-15
-5
-4
-3
-2
Real Axis

-1
Desired eigen value is the intersection of the

RL plot with constant - line.
s1 = 1.05 2.41 j
s = 2.9
s2 = 2.16 4.96 j
s = 0.68
K ( ) s=s = 1
1
for s1 : K = 8.98 or k = 0.45
for s2 : K = 28.26 or k = 1.43

INTRODUCTION TO MATLAB
MATLAB commands used
rlocus
rltool
Use MATLAB help for further information.

EXAMPLE
ELECTROMAGNETIC RELAY CONTROL
(Refer Handout for figure)
( s )
=
OLTF : G ( s ) =
s 3 + 16s 2 + 44 s 160 Vin ( s )
1
Controller
R
Vin
Plant: Electromagnetic
Relay (3rd Order)

CE : s 3 + 16s 2 + 44s 160 + k = 0

Stable or not? and why?
RH : k 160 > 0
Root Locus
20
n=3
na = 3
a = 60(2q + 1)
a = 5.33
10
Imaginary Axis
m=0
15
5
0
-5
-10
-15
-20
-30
-25
-20
-15
-10
-5
Real Axis

10
At intersection with Im Axis s = j

CE : ( j )3 + 16( j )2 + 44( j ) 160 + k = 0
j 3 16 2 + 44 j 160 + k = 0
(16 2 160 + k )+ ( 3 + 44 )j = 0
( 2 + 44) = 0 = 0 or = 44
16 2 160 + k = 0
16(44 ) 160 + k = 0
k = 864

RH : Let = k 160
s 3 + 16 s 2 + 44s + = 0
1 1
b1 =
16 16
1
= ( 704)
16
44
s3
44
s2
s1
b1
k < 864
s0
c1
c1 = = k 160 > 0 k > 160
b1 > 0 < 704 k 160 < 704
160 < k < 864

FREQUENCY RESPONSE METHODS

Frequency Response: It is the steady state response
of a system due to a sinusoidal input for a stable or
marginally stable system.
Why use frequency response methods?
For LTI system, the response is of the same form as
the input. If input is sinusoidal (w, A), output will
also be sinusoidal (w, B, ).

INPUT/OUTPUT RELATIONSHIP FOR F.R.

y (t )
x(t )
STABLE
G (s )
L.T.I.
Y (s)
X (s)
L
Let : x(t ) = X sin t X ( s ) =
Xs
s2 + 2
p( s)
p(s)
=
G ( s) =
q ( s ) (s + s1 )(s + s2 )......(s + sn )
Y ( s) = G (s) X ( s)
p(s)
p(s)
Xs
=
X (s) =
(s + s1 )(s + s2 )......(s + sn ) s 2 + 2
q(s)

Partial Fraction Expansion

bn
b1
b2
a
a
Y ( s) =
+
+ ... +
+
+
s + s1 s + s2
s + s n s + j s j
a = conjugate of a
L- 1
y (t ) = b1e s1t + b2e s2t + ... + bn e sn t + ae jt + a e jt

1444442444443
= 0 for steady state (t 4 )
yss (t ) = ae jt + a e jt

Residue theorem a & a

a = Y ( s ) ( s + j ) s = j
X
(s + j ) s = j
a = G( s)
(s + j )(s j )
X
a = G ( j )
2 j
X
& a = G ( j )
2 j
G ( j ) = G ( j ) e j
= G ( j )
G ( j ) = G ( j ) e j

a=
G ( j ) Xe j
y ss (t ) =
2j
& a=
X G ( j ) e j e j t
2j
G ( j ) Xe j
2j
+
X G ( j ) e j e j t
2j
e j ( +t ) e j ( +t )
+
y ss (t ) = X G ( j )
2 j
2 j
e i e i
We know , Euler' s identity sin =
2j

y ss (t ) = X G ( j ) sin(t + )
> 0, phase lead

< 0, phase lag
EXAMPLE
x(t )
G (s )
k
G(s) =
s +1
y (t )
x(t ) = X sin t
y ss (t ) = X G ( j ) sin(t + )

k
k
G ( j ) =
G ( j ) =
1 + j
1 + 2 2
= G ( j ) = k (1 + j )

= atan
1
yss (t ) = X
k
1 + 2 2
sin (t atan ( ) )

OBSERVATIONS :
<< 1 ( very small)

Magnitude of s.s. is almost kX
Phase of s.s. is very small
>> 1 ( very large)

Magnitude of s.s. is approximately 1
Phase of s.s. is approximately 90o

BODE PLOT
Amplitude
& Phase
Linear
MAGNITUDE
Logarithmic w
Linear
PHASE
Logarithmic winput

Bode Plot:
It is a graphical representation of the steady state
output due to a sinusoidal input for a range of values
of the input frequency.
Standard Representation for Magnitude of G(jw) is
logarithmic and is given as
20 log10 G ( j ) db (decibels)

How to Draw a Bode Plot
In General, a T.F.
G ( s) =
k g (s + a )
2
2
(
(
)
+
+
+
2
s s b s
n s n )
Normalizing,
1
k g a s + 1
a
G ( s) =
2
2
1
2 s
1
s b s + 1 n
s
+
+
n
n

s = j
k ( aj + 1)
G ( j ) =
j 2 2
+
j ( bj + 1)
j + 1
n
n
G ( j ) =
k (1 + aj )
2 2
j (1 + bj ) 1 +
n
n
Plot Magnitude in db

20 log G ( j ) = 20 log k + 20 log (1 + aj )
20 log j 20 log (1 + bj )
2 2
+
20 log 1
n
n
20 log G ( j ) = 20 log k + 20 log 1 + ( a )2

20 log 20 log
2
(1 + ( b )2 )
2 2
2
2

+
20 log 1
n n

Phase
G ( j ) = k + (1 + aj ) j
2 2
+
(1 + bj ) 1
n
n
a

= + a tan
atan
0
1
2
n
b
a tan
a tan
2
1
1
n

Constant
20 log k constant in db
k = 0
Linear Axis
| |
db
0.1
10
o
0.1
10

Pole at Zero or Origin

1
1
Pole at origin
s
j
1
= 20 log j = 20 log 2
20log
j
= 20 log db
= atan = 90o
0
j
(Constant )
0.1
20 log
20
20 log
20
1
10
0
20
0
20

40
Passes Through
=1
20
| |
db
0.1
10
100
(log)
-20
Slope -20 db/dec
-40
1
2
1 j
j
Slope -40 db/dec
o
0.1
-90
-180
10
100
1
j
2
1
j

(log)
Dual Pole at Origin
20 log
= 20 log
( )
= 20 log
2 2
= 20 log 2 = 2( 20 log )
0
= 1 = 0 atan
2
2
1
( )
= 180o = 2 90o


1

N 20 log
SN
j
where N slope 20 N db/dec

N ( 90 )
S N Mag passes through = 1
Slope N 20 db/dec
Phase N 90o

Simple Pole or Zero
1
1 + j
(1 j )
Break Frequency
Corner Frequency
1
2 2
20 log
= 20 log 1 +
1 + j
= 10 log 1 + 2 2
when =
= 1 2 2 = 1
= 10 log(1 + 1) = 10 log 2 = 3.01 db

<<
1
42
4
31
<< 1
= 1
>>1
Mag 10 log 1 = 0 db
0o
Mag 3.01 db
45o
Mag 10 log 2 2
20 log( )
20 db/dec
90o

0.1
| |
db
10 1
100 1
(log)
-20
Slope -20 db/dec
-40
0.1
10 1
100 1
-45
-90

(log)
(1 + j ) N
Mag : << 1 0 db
= 1
3.01 db ( N )
>> 1 20 ( N ) db/dec
Phase : << 1 0o
= 1
45 ( N )
>> 1 90 ( N )

Quadratic Poles & Zeros

Quadratic Poles
1
G ( s) =
s 2 + 2 n s + n2
G ( s) =
n2 s 2
s
+ 2
+1
n
n
G ( j ) =
1
2
+ 2
1
j
n
n

+ 2
j
Mag : 2 log G ( j ) = 20 log 1
n
n
2 2
2
1

+ 2
= 20 log 1

2
n n
Let
=u
n
2
2 2
= 10 log 1 u
+ (2u )

2u
Phase G ( j ) = atan
2
1 u
if u << 1, Mag 10 log 1 = 0
Phase 0o
if u >> 1, Mag 10 log u 4 = 40 log u
slope = 40 db/dec
2u
1
atan
Phase -atan
2
u
u
180o

if u = 1, Mag = 10 log(2 )2 = 20 log(2 )

2
Phase = atan
0
| |
db -20
0.1u
u =1
o
90
=
10u
-40
(log)
Slope -40 db/dec
0.1u
u =1
10u
(log)
-90
-180

Max Amplitude occurs at r (resonant frequency)
r = n 1 2 2
< 0.707
M P r = G ( jr ) = 2 1 2
Mag
G ( j ) sin t + 0o
Output =
Input

First Order Simple Pole

at = 0.1 0 db
20 log G ( j ) = 0 db G ( j ) = 1
mag
1 sin(t + 0o )
output =
input
at = 100
20 log G ( j ) = 40 db G ( j ) = 0.01
90
mag
0.01 sin(t 90o )
output =
input
Example
G ( s) =
10 ( s + 3)
s ( s + 2) ( s 2 + s + 2)
Normalize
1
10 3 s + 1
3
G (s) =
2
1
1
s
s 2 s + 1 2
+ s + 1
2
2
2
j
7.5
+ 1
3
G ( j ) =
2
j
j j
j
+1
+ 1
+
2
2
2
FACTORS
Break Freq.
Slopes
db/dec
Angle
deg.
7.5
+1
( j )

j + 1
2
(Quad )1
Low
Freq.
-20
High
Freq.
20
-20
-20
-40
Low
Freq.
-90
-45
-90
-90
-180
Break
Freq.
High
Freq.
45
0
90
-90

Final Slope ( >> 1) = 60 db/dec

Final Phase ( >> 1) = 270o
Constant :
20 log(Constant) db =
20 log(7.5) = 17.5 db
Quadratic :
Find r , , M p r from formulae.

Discussion/review from last example
| |
db
-20
-40
Simple Zero
Slope = 20 db/dec
Constant
r = M p
Simple Pole
Slope =
-20 db/dec

(log)
Quadratic
Slope =
-40 db/dec
Magnitude
| |
db
-20
-40

(log)
Phase
90
45
(log)
o
-45
-90
-135
-180
-270

MINIMUM PHASE & NON-MINIMUM

PHASE SYSTEMS
Poles & Zeros in left-hand s-plane
Minimum Phase
Poles or Zeros in right-hand s-plane
Non-Minimum Phase
1 + j T
1 j T
G1 ( j ) =
G2 ( j ) =
1 + jT1
1 + jT1
T
T1
G1 = atan
atan
1
1
T
T1
G2 = atan
atan
1
1

1 1
>
T T1
o
o
1
T1
1
T1
1
T
1
T
CUT-OFF FREQUENCY OR BANDWIDTH

Cut-off frequency is the frequency at which the
magnitude of the closed loop frequency is at 3 db
below the zero frequency value.

| |
db
(log)
-3 db
Bandwidth
| |
db
(log)
0
-3 db
Bandwidth
Find the attenuation factor at -3 db magnitude.

(Find real # corresponding to magnitude of -3 db)

e j L
G ( j ) =
1 + j T
L = 0.5 T = 1.0
Magnitude (db) = 20 log G ( j )

= 20 log e jL 20 log 1 + jT
Phase = e jL 1 + jT
T
=
L a tan
{
1
57.3L deg
= L or 57.3L

STABILITY ANALYSIS IN FREQUENCY

DOMAIN USING BODE PLOTS
cg
| |
db 0
o
-180
(log)
Positive Gain
Margin (G.M.)
Positive Phase
Margin (P.M.)
cp

(log)
Phase Margin: It is the amount of additional phase

lag at cg required to bring the system to the verge
of instability.
= 180 +
cg
(
)
if < 0 ( < 180o ) , System is unstable
if > 0 > 180o , System is stable
Gain Margin: It is the reciprocal of the magnitude

|G(j)| at cp. GM indicates how much the gain can
be increased before the system becomes unstable
OR how much it must be decreased for the system to
become stable.

kg =
1
G ( j )
cp
k g db = 20 log k g = 20 log G ( j )
cp
if k g (db) > 0, system is stable

if k g (db) < 0, system is unstable
For satisfactory performance and to guard against
variations in performance of system components,
30o < PM < 60o

GM > 6 db

Example
k
G (s) =
s (s + 1)(s + 5)
Find k for stability and draw BP for k = 1,10,100

Root Locus
10
n=3
k = 30
m=0
= 5
na = n m = 3
a = 60(2q + 1)
6
a =
= 2
3
Imaginary Axis
-0.472
-5
-10
-14
-12
-10
-8
-6
-4
Real Axis

-2
STATIC ERROR CONSTANTS

Slope of magnitude curve at low frequencies
G( s) =
Type 0
k (T1s + 1)(T2 s + 1)....
s N (Ta s + 1)(Tb s + 1)....
(s = j )
Static position
At low frequencies, the magnitude of G(j) equals to

k or kp
lim G ( j ) = k = k p

kp db
| |
db
-20 db/dec
-40 db/dec
(log)
Type 1 Static Velocity
k (s + 2 )
G( s) =
s (s + 5)(s + 8)
| |
db
k ( j + 2 )
G ( j ) =
j ( j + 5)( j + 8)

(log)
At low frequencies,
kv
, << 1 | |
G ( j ) =
j
db
kv = 1
-20 db/dec
20logkv
0
1 (log)
=1
Type 2 Static Acceleration

At low frequencies, the slope is -40 db/dec

G ( j ) =
20 log
ka
( j )2
ka
( j )2
ka
( j )
ka
= 20 log 1
=1
=1
, << 1
-40 db/dec
| |
db
20logka
0
(log)
=1
a = k a

a = k a
COMPENSATION
Series/Cascade
R
Feedback
Output/Load
Input
Gp
Gc
H
Gc
Gp
H
P, PI, PD, PID, VF


LEAD / LAG COMPENSATORS
s +
G( s) = k
s+
Lead:
Relates to phase lead
Lag:
Relates to phase lag
Lead-Lag:
Low frequencies Lag

High frequencies Lead

Lead: - Improves transient response

- Small Effect if ess
- Accentuates high frequency noise effects
Lag: - Increases transient response time
- Appreciable improvement in ess steady state
accuracy
- Suppresses high frequency noise effects
Lead-Lag: - Combines characteristics of both
(diff. freq. range)
- Increases system order by 2.

LEAD COMPENSATORS
1
s+ 1
1 s+ T
T
G (s) =
=
s+ 1
11
T s +
T
<1
1
s
+
1
T
=
s + 1T
LEAD
Pole to the left of zero
>1
1

| |
db 0
LEAD
s+ 1
T
G ( s ) = kc
s+ 1
T
90
45
o 0
- 45
- 90
0 < <1
STATIC / DC
GAIN
kc
min = 0.07
Maximum Phase Lead = 60o

Find Maximum Phase Lead

Max. value occurs at MEAN, geometric mean
between pole and zero
Is the halfway distance on the log plot
1 1
1
m = zp =
=
T T T
T j + 1 T j + 1
G ( j ) = kc
Tj + 1 Tj + 1
G ( j ) =
(
1 + T 2 2 + jT (1 ))
(T )2 + 1
kc

1
max =
= atan
2
T
1
1
sin max =
1+
HOW TO DESIGN A LEAD COMPENSATOR
BASED ON FREQUENCY APPROACH
R
Gp
4
Gp =
s (s + 2 )

Design a Compensator such that

static velocity error const. kv=20 sec-1
phase margin at least 50o
gain margin at least 10 db
s+ 1
Ts + 1
T
= kc
G ( s ) = kc
Ts + 1
s+ 1
T
0 < <1
OLTF compensated system
4
Ts + 1
GcG p = kc
Ts + 1 s (s + 2 )

Let kc = k
1
Ts + 1
GcG p = 4k
(Ts + 1) s(s + 2)
1
Let 4k
= G1
s (s + 2 )
Ts + 1
GcG p = G1
(Ts + 1)

4k
kv = 20
kv = lim sGcG p
s 0
Ts + 1
1
4k
= lim s
s (s + 2)
s 0 Ts + 1
kv = 2k k = 10 = kc

k = 10
1
Bode Plot : 4k
s (s + 2 )
Examine new Bode Plot with k = 10
kv = 20 sec-1 (See Bode Plot: Draw a line extension
of the low frequency 20 db/dec part of the
magnitude plot and find its intersection with the 0
db line)
Find the new phase and gain margins.

GM =
PM = 180 162 = 18o
Requirement for PM 50
Therefore, need to add phase (without changing k)
Amount of phase needed = 50 18 =32
Need a LEAD compensator
Lead compensator will shift gain crossover
frequency to the right, thus reducing the PM.
So, require the compensator to provide additional
phase of (say) 5.

Therefore, the total phase lead needed,

32 + 5 = 37
Determine the frequency at which the magnitude of
1
the uncompensated system G1 equals 20 log
Select this frequency to be the new gain crossover

frequency.
Determine the attenuation factor
1
sin =
Solve for
1+

Identify frequency at which max phase lead occurs

1
max =
T
The frequency corresponding to 40 PM is about 0.7
rad/sec. The new gain crossover frequency must be
chosen near this frequency.
1
( zero of lag compensator)
=
T
Lets select this to be 0.1 rad/sec (not far below 0.7 ??
small modification on phase plot)

G1 ( j ) = 20 log
1
= 20 log
= 6.2 db
0.24
Refer Bode Plot and find for G1 ( j ) = 6.2db
= 9 rad/sec
Determine the corner frequencies for L.C.
zero
1
z =
T
&
pole
1
p =
T

cg = 9 =
1
T
1
= 4.41 &
T
T = 0.23
1
= 18.37
T
L.C.
s+ 1
s + 4.41
T
Gc = kc
= kc
s + 18.37
s+ 1
T
0.23s + 1
or Gc = kc
0.054 s + 1

10
k = kc kc = =
= 41.7
0.24
k
s + 4.41
Gc ( s ) = 41.7
s + 18.37
1
G p ( s) = 4
s (s + 2 )

EXAMPLE: Position Control System

Preamplifier
Shaft
Motor
Amplifier & Load velocity
Compensator
1/s
100
s + 100
1
s + 36
Design a lead compensator to yield a 20 % overshoot

kv =40 sec-1, peak time Tp = 0.1 sec
Solution:
Closed loop bandwidth to meet the speed of response
requirement imposed by Tp=0.1 sec

BW =
Tp 1 2
1 2 2 + 4 4 4 2 + 2
is found from overshoot requirement, for 20%
= 0.456
BW = 46.59 rad
Meeting kv requirement, need to increase the
constant gain, gain = 1440
Plot the uncompensated gain bode plot. Find PM
and GM.

20% overshoot requires
m = tan
2
2 2 + 1 + 4 4
= 48.15
s + 25.27
G ( s ) = 2.38
s + 60.17

LAG COMPENSATION
1
+
s
Ts + 1
T
= kc
Gc ( s ) = kc
Ts + 1
s+ 1
T
| |
0
zero : 1
db
T
pole : 1
o 0
T
>1
1
T

Example
1
Gp =
s(s + 1)(0.5s + 1)
Gp
Static velocity error constant kv = 5 sec-1

PM at least 40
GM at least 10 db
Plot BP
Solution
g = 0.75 rad GM = 9.5 db

cp = 1.4 rad PM = 32
kv = 1sec 1

Ts + 1
Gc ( s ) = kc
Ts + 1
>1
G1 ( s ) = kG p ( s ) = adjust gain k to meet kv req.

kv = lim sG ( s ) G p ( s )
s 0
Ts + 1
1
= lim s
k
s 0 Ts + 1 s (s + 1)(0.5s + 1)
kv = k = 5
Plot new BP including k = 5

cg = 1.8
GM = 4.4 db
cp = 1.4
GM = 13
Satisfying kv requirement made the system unstable

Desired phase is " padded" by about 10 - 12
Required phase is 52
Phase angle for uncompensated system is - 128
at = 0.5 rad/sec
Select this as the new gain crossover frequency
Mag plot = 0 db at = 0.5

The attentuation to bring the magnitude plot to 0 db

1 attenuation to bring
20 log =
mag. plot to 0 db
1
20 log = 20 db
= 10
5
k
k = kc kc =
=
= 0.5
10
1
Plot of lag compensator =
=1
T

zero = 0.1
pole = 0.01
kc = 0.5
s+ 1
Ts + 1
T
= kc
Gc ( s ) = kc
Ts + 1
s+ 1
T
s + 0.1
= 0.5
s + 0.01

PHYSICAL REALIZATION OF
CONTROLLERS & COMPENSATORS
Ro
Ri
Vo
Vi
Ro
Vo ( s ) = Vi
Ri
Proportional Gain

Zo
Zi
Vi
Vo
Z: Impedence
Zo
Vo ( s ) = Vi
Zi
R2
Vi
R1
C2
Vo
Zo
Vo ( s ) = Vi
Zi

R1
R2
R
2
Gain
Int.
Diff.
PI
R
C
PD
RC
s
RCs
R2
s+ 1
R
R2C
2
R1
s
R2C s + 1
R1C
R1
R1

C1
R2
PID
C2
R1
i 1
R2 C1
R1C2
+ R2C1s +

+
R1 C2
s
kp

LAG/
LEAD
s+ 1
C1
R1C1
1
C2 s +
R
C
2 2
C1
C2
R1
R2
LAG
LEAD
R2C2 > R1C1

R1C1 > R2C2

Passive Components
LAG
Vi
R1
Vo
R2
C
s+ 1
Vo ( s )
R2
=
Vi ( s ) R1 + R2 s + 1
R2C
(R1 + R2 )C

LEAD
C
Vi
R2 V o
R1
s+ 1
Vo ( s )
=
Vi ( s ) s + 1
R1C1
+ 1
R1C
R2C

LAG-LEAD
C1
Vi
Vo ( s )
=
Vi ( s )
s 2 +
R1
R2 V o
C2
s + 1
s + 1
R
C
R
C
1 1
2 2
s+ 1
1
+ 1
+ 1
R1C1
R2C2
R2C1
R1R2C1C2

Example: Implementation of PID

(
s + 55.91)(s + 0.5)
Gc ( s ) =
s
27.95
Gc ( s ) = s + 56.41 +
s
From Realization table,
R2 C1
= 56.41
+
R1 C2
1
= 27.95
R1C2
R2C1 = 1

3 equations, 4 unknowns.
Let C2 = 0.1F C1 = 5.59 F
R1 = 357.73 k
R2 = 178.86 k
R2=179 k
Vi
C1= 5.6 F
C2= 0.1 F
R1= 358 k

Vo
Determination of OLTF from Experimental Data

Sensor
Shaker Table
Sensor measures displacement

Shaker table generates
sinusoidal signal of certain
Amplitude & frequency
Turn displacement into dB

Plot Bode Plot
Determine if sys is Min/Non-Min phase

examine high frequencies
slope ~ -60dB/dec relative order -3
phase = -270 relative order -3
Minimum phase

Identify quadratic terms

establish natural frequency for quadratics
at = 180 = 2 rad/sec
max. amplitude ( = 2 ) ~ 14 db
establish n = 2
M p = 14 db 20 log M p = 14 db
M p = 5.02
Mp =
1
2 1 2
= 0.1

Determine corner frequencies

examine slope changes. Corner freq at w = 0.5
for first order numerator.
corner
+ 1 corner = 0.5
Low Frequencies
Establish double pole at origin
ka = 1
The extension of -40 dB/dec (low

freq). Slope intersects the 0 dB line
at w = 1.

Set-up T.F. Relative Order = -3
G ( j ) =
j
+ 1
c
2
( j )2 j + 2
j + 1
n
n
j
+
1
0.5
2
2
( j ) j + 2(0.1) j + 1
2
2

G (s) =
8(s + 0.5)
s 2 s 2 + 0.4 s + 4

TUNING RULES FOR PID CONTR.

Ziegler-Nichols Rules
First Method:
Obtain unit step response of system if unit step
response looks like an S-shaped curve
C(t)
Inflection point
Tangent at inflection point
t
Tunit
step
Plant

Use L & T to define the gains of PID controller
1
Gc ( s ) = k p 1 +
+ d s
is
Type of
controller
kp
T/L
PI
0.9 T/L
L/0.3
PID
1.2 T/L
2L
0.5L

PID Controller
T
1
+ 0.5 Ls
Gc ( s ) = 1.2 1 +
L 2 Ls
s +
L
= 0.6T
s
Im
L
Re

Second Method:
Controller
r(t)
Plant
c(t)
Assume that ( i = & d = 0 )

Only proportional kp
Start increasing kp from 0 critical value kcr s.t.
the output sustains oscillations

Pcr
C(t)
Type of
controller
kp
0.5kcr
0
0
0.125Pcr
PI
0.45kcr
1
Pcr
1.2
PID
0.6kcr
0.5Pcr

PID Controller
1
+ 0.125 Pcr s
Gc ( s ) = 0.6kcr 1 +
0.5Pcr s
4
s +
Pcr
= 0.075kcr Pcr
s

Application of Z-N:
r(t)
Gc
Gp
c(t)
Desired ~ 25% overshoot

1
G p (s) =
s (s + 1)(s + 5) Use 2nd Z-N start with kp
kp
C
=
R s (s + 1)(s + 5) + k p
RH : s 3 + 6 s 2 + 5s + k p = 0
kcr = 30

Identify Pcr (critical period)
Period =
CE : s 3 + 6 s 2 + 5s + 30 = 0
2
CE s = j = 5 Pcr =
= 2.81
5
Use table to find k p , i , d
k p = 0.6, kcr = 18, i = 0.5, Pcr = 1.41,
d = 0.125, Pcr = 0.35


Gc
(
s + 1.42)2
( s ) = 6.32
s
Experiment with double zero at s = 0.65

s = 1.4235

ANALYSIS OF CONTROL SYSTEM IN

STATE SPACE REPRESENTATION
Assume that system is representation in terms of nfirst order dfqs
Preferred representations
Canonical Forms
Controllable, Observable, Diagonal, Jordan
Controllable important when discussing poleplacement approach for control system design
MATLAB: ctrb(a,b)
obsv(a,c)

Example:
Controllable:
Observable:
Y (s)
s+3
=
U ( s ) s 2 + 3s + 1
x&1 0
=
x&2 2
x1
y = [3 1] + 0
x2
x&1 0 2 x1 3
+ u
=
x&2 1 3 x2 1
y = [0
1 x1 0
+ u
3 x2 1
x1
1] + 0
x2

Controllability: A system is said to be controllable at

time to if it is possible by means of an unconstrained
control vector to transfer the system from an initial
state x(to) to any other state in a finite time interval.
x(to) x(tf)
Observability: A system is said to be observable at
time to if, with the system at state x(to) it is possible
to determine this state from the observation of the
output over a finite time interval.

Observability:
C
CA
Q = CA2
M
CAn 1
Find det Q if Q is rank n
Controllability:
Pc = B
AB
A2 B
An 1B
if det Pc 0 system is controllable


Example:
x&1 3
=
x&2 2
y = [1
1 x1 0
+ u
1.5 x2 1
x1
1]
x2
Controllability: Pc = [B
1
AB ] =
4
det Pc = 0
X1( s)
s + 2.5
=
U ( s ) (s + 2.5)(s 1)
pole/zero cancellation

Observability:
1
Q=
5
1
det Q = 2.5 + 5 = 7.5 0
2.5
System observable.

Controllability / Observability and Pole / Zero

Cancellation
1
x& =
0
0
b1
x + u
2
b2
y = (c1
c2 ) x
y = c(sI A)1 Bu
y (b1c1 + b2c2 ) s (2b1c1 + b2c2 )
=
(s 1)(s 2)
u
b2c2 s b2c2
b2c2 (s 1)
Let b1 = 0 TF =
=
(s 1)(s 2) (s 1)(s 2)

Let b2 = 0 Controllability
c1 = 0
[B
AB ]
c2 = 0
b1
b2
b1
b
2
b1
2b1b2 b1b2 = b1b2
2b2
b1

b2
Observability
C1
C
CA = (C
1 C2 )
C2
C1
0 =
C1
2

C2
2C2
det = 2C1C2 C1C2 = C1C2

Pole Placement
Define the location of the closed loop poles
Estimate the controller gains to yield the defined
C.L. poles.
All system states are successfully measured.
If a state is non-measurable or non-available, then
we can implement an observer to estimate the state.
x& = Ax + Bu
y = Cx + Du
Assume
Controllability
Select control signal u = kx


x& = Ax Bkx
x& = ( A Bk )x
(
A Bk )t
x(t ) = e
x(0)
Stability and transient

performance determined from
eigen values of (A-Bk)
Generate the CE for (A-Bk);

Generate the CE for desired pole locations. Equate
the two.
Evaluate k through direct substitution.
Low order system, n = 3
k = (k1
k2
k3 )
Desired eigen values are 1, 2 , 3


CE : (s 1 )(s 2 )(s 3 ) = 0
CE from A Bk
sI ( A Bk ) = 0
Equate the two CEs
Ackermans Formula or Principle
x& = Ax + Bu
x& = ( A Bk )x
u = kx
k = [0
Controllab ility matrix
1] B
AB K A B ( A)
14444244443
n 1

( A) : State transition matrix

( A) = An + 1 An 1 + 2 An 2 + K + n 1 A + I
CE of new system with k
sI ( A Bk ) = s n + a1s n 1 + a2 s n 2 + K + an
Example
0
A = 0
1
1
0
5
0
1
6
0

B = 0
1

Desired C.L. poles : s = 2 4 j

s = 10. Find k .

Check for controllability B
AB
A2 B
Find det = 1 state controllable

Let k = [k1
k2
k3 ]
CE of new system ( A Bk )
sI ( A Bk ) = s 3 + (6 + k3 )s 2 + (5 + k 2 )s + 1 + k1 = 0
CE based on desired eigen value loc.
s 3 + 14 s 2 + 60 s + 200 = 0
Equate coefficients

6 + k3 = 14 k3 = 8
5 + k 2 = 60 k 2 = 55
1 + k1 = 200 k1 = 199
Ackerman's
k = [0
1] B
AB
A2 B ( A)
( A) = A + 14 A + 60 A + 200 I =
After all algebra k = [199
55
8]

OBSERVER
x& = Ax + Bu
y = Cx
Observer x& = Ax + Bu + L( y Cx )
as t , x x
x(to ) is not known
x (to ) estimate
Observer error, e(t ) = x(t ) x (t )
We want the observer to produce a result s.t.
e(t) 0 as t

If the system is completely observable we can

find L so that the tracking error e(t) is asymptotically
stable
e& = x& x&
= Ax + Bu Ax Bu L( y Cx )
= A( x x ) L( y Cx )
e& = Ax Ax Ly + LCx Roots or eigen values
= A( x x ) LC ( x x ) should be in the negative
s-plane
&e = ( A LC )( x x )
e& = ( A LC ) e
det sI ( A LC ) = 0


Classical Methods of Control Systems (Shiakolas)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Classical Methods of Control Systems (Shiakolas)

Uploaded by

Copyright:

Available Formats

BLOCK DIAGRAM REPRESENTATION

ml 2&& + mgl sin =

ml 2&& = mgl sin

ME 5303 Classical Methods of Control

LINEARIZATION Taylors Series Expansion

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

If the equilibrium point is = 180o

ME 5303 Classical Methods of Control

LAPLACE TRANSFORM (Assume Zero ICs)

ME 5303 Classical Methods of Control

C.E. for 0 = 180o

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

Set of first order DFQs

ME 5303 Classical Methods of Control

A = coefficient matrix square

C.E. det (SI A) = 0

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

Derive system dynamic equations

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

TRANSFER FUNCTIONS (T.F.)

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

PRINCIPLE OF SUPERPOSITION (LTI)

ME 5303 Classical Methods of Control

STATE SPACE REPRESENTATION

ME 5303 Classical Methods of Control

2 nd order linear DFQ 2 first order linear

ME 5303 Classical Methods of Control

Only one variable for system state

ME 5303 Classical Methods of Control

Y = C (sI A)1 B + D U + C (sI A)1 x0

Denominator of the above eq. sI A

ME 5303 Classical Methods of Control

Time Step for Numerical Integration

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

(s 2 + 2s )(s + 1) + 10 = 0 3rd Order

ME 5303 Classical Methods of Control

Transfer Function has zeros or derivatives of input

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

State space representation w.r.t. variable z

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

FEEDBACK CONTROL SYSTEM

ME 5303 Classical Methods of Control

Find such that

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

GENERAL PROPERTIES OF F.B.C.S.

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

SENSITIVITY TO SENSOR GAIN

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control

= 0.6 & n = 5 rad/sec

ME 5303 Classical Methods of Control

ME 5303 Classical Methods of Control