Lecture on Controller Design

All control systems are designed to achieve specific objectives.

Design process normally involves the following procedure:

• Specify what system should do, i.e what are the specifications to
be fulfilled.
• Determine controller configuration/types of compensations.
• Determine the controller parameters for the controller structure

Specifications
• Normally consists of relative stability, transient response, steady
state response and frequency response.
• The sets of design specifications are normally expressed as rise
time, delay time, settling time, maximum overshoot, maximum
overshoot time, steady state error etc.
• In practice, if a system is to be designed as to meet the
required specifications, therefore, it is necessary to alter
the system by adding an external device to it.

• Such a redesign or alteration of system using an

additional suitable device is called compensation of a
control system.

• The external device – used to alter the behavior of the

system to achieve the desire specification is called
compensator.
 Types of compensation

Basic block diagram

r t  e t  u t  y t 

 Controller Plant


Series / cascade Compensation

The compensator is a physical device whose transfer function is

denoted as Gc(s).

If the compensator is placed in series with the forward path transfer

function of the plant, the scheme is called series compensation.
The flow of signal in such a series scheme is from lower energy level
towards higher energy level.
 r t   e t   u t  y t 
  Plant

 
Controller

Parallel / feedback compensation

In some cases, the feedback is taken from some internal element and
compensator is introduced in such a feedback path to provide an
additional internal feedback-loop is called feedback compensation.
The energy transfer in such feedback scheme is from higher energy
level towards lower energy level point. Thus, the number components
required to design are less than series compensation.
 r t  e t  u t  y t 
 
Controller
 Plant
Gc  s 
 

Controller
GH  s 

Series-Feedback Compensation

In some cases, it is necessary to provide both types of

compensations, series as well as feed back
r t  Controller  Controller u t  y t 
 Plant
GH  s  Gc  s 

Series-Forward Compensation
 Compensators
Compensators that use pure integration for improving steady-state error or
pure differentiation for improving transient response are defined as ideal
compensators.

• Ideal compensators must be implemented with active networks. E.g. active

amplifiers and possible additional power sources.

• An advantage of ideal integral compensator is that steady state error is

reduced to zero.

• Ideal integral compensator is called a proportional-plus-integral (PI)

controller. Non-ideal integral compensator is called lag compensator.

• Ideal derivative compensator is called proportional-plus-derivative (PD)

controller. Non-ideal derivative compensator is called lead compensator
 Selection of compensator

(1) Flow
• Very noisy, thus need derivative-action
• Overall gain less than, integral-action will ensure no steady-state error.

(2) Level
• Normally is of type 1, thus just required proportional action

(3) Temperature
• Thermal delay, thus normally required PID-action
(4) Pressure

• Characteristic can be fast or slow depending on application, thus required PI or P.

 P-I-D compensator

Proportional compensator (P)

Use to improve steady state error type

R(s) + E(s) Y(s)

Gc A(s) Gp
-
B(s)
H (s )

Consider P-compensator transfer function as Gc  K p and A( s )  K P E ( s )

where A(s) is the actuator signal.
• High K p gives better steady state but poor transient response

 Too high K p can cause instability.

Example: Find the steady state error

 i (s ) + Kp 1 (s)
- Js  F

Closed-loop transfer function

1
KP.
( s ) Js  F  KP

 i (s) 1 Js  F  K P
1 KP.
Js  F
For a unit step input
KP 1
( s ) 
Js  F  K P s
Final value
KP 1 KP
ess  lim s 
s 0 Js  F  K P s F  K P
 Integral compensator (I)

Use to eliminate error for type 0

K I and actuating signal of K
Consider the I-compensator Gc  A( s )  I E ( s)
s s

R(s) + E(s) A(s) Y(s)

Gc Gp
-
B(s)
H (s )

 Slow response, can be used with P-compensator to remedy this problem

Example: Find the steady state error

 i (s ) 1 (s )
KI
s Js  F
+ -

KI 1
.
( s ) s Js  F  KI
Closed-loop transfer function 
 i (s) KI 1 Js 2  sF  K I
1 .
s Js  F
For a unit step input, the response is
KI 1
( s ) 
Js 2  sF  K I s

KI 1 KI
ess  had s 2
 1
s 0 Js  sF  K I s K I
 Derivative compensator (D)

Derivative compensator (D)

Consider the D-compenator as Gc  sK D and actuating signal as A( s)  sK D E (s)

R(s) + E(s) A(s) Y(s)

Gc Gp
-
B(s)
H (s )

 Quick response
 No effect at steady state because no error signal
 Useful for controlling type 2 together with a P-controller
 Response only to rate of change and no effect to steady
state
  o (s) km
Example: Determine kd , if damping ratio 0.6 and k a  0.2 , Gm ( s )  
va ( s ) s m s  1
Error detector Amplifier Servo motor
r va o
ke e ka Gm (s )
,,

where k e  5 V.rad
-1
k m  20 rad.s 1 .V 1 and  m  0.05 s

Compensator
Va ( s )
 k a (1  sk d )
E (s)

Open loop Transfer function

km kkk
Go ( s)  k e .k a . .1  e a m
s m s  1 s m s  1
 New open loop transfer function

k e k a (1  sk d )k m
Go ( s ) 
s  m s  1
Closed loop transfer function

k e k a (1  sk d )k m
Gc ( s ) 
s m s  1  k e k a (1  sk d )k m

Substitute in closed loop TF

. , k e  5 V .rad -1 , k m  20 rad.s 1 .V 1 ,  m  0.05 s ,  m  01
k a  01 . s   0.6 and . k a  0.2

20(1  sk d ) 400(1  sk d )
Gc ( s )   2
s 0.05s  1  20(1  sk d ) s  20s (1  20k d )  400

Do comparison with general second order characteristic equation

 n  400  20 rad .s 1 and 2  0.6  20  20(1  20k d )

Hence
2  0.6  1
kd   0.01
20
Proportional-integral compensator (PI)

Use for combination of fast convergence zero steady state error

KI
Gc ( s )  K P  or
s

 1 
Gc ( s )  K P 1   Where Kp/KI is called the integral constant, TI
 T s
I 
Example: Find the steady state error

Consider a speed control system employing a PI-controller

 i (s ) K 1 (s)
KP  I Js  F
+ - s

Closed-loop transfer function

 K  1 
 K P  I . 
( s )  s   Js  F  sK P  K I
  2
 i (s)  K   1  Js  s( F  K P )  K I
1   K P  I . 
 s   Js  F 

For unit step input

sK P  K I 1
( s )  .
Js 2  s ( F  K P )  K I s

KI
 ss  had s .( s )  1
s 0 KI
Proportional-integral-derivative
compensator (PID)

Ki
Gc ( s )  K p   sK d or
s

 1 
Gc ( s)  K p 1   sTd 
 Ti s 
 Characteristics of Phase-Lead and
Phase-Lag Compensators
Phase-Lead Compensator (PD type compensation)

1. Increases system bandwidth which usually correlates to reduced

rise and settling times and a susceptibility to high frequency noise.
2. Increases gain at higher frequencies.

Phase-Lag Compensator (PI type compensation)

1. Decrease system bandwidth
2. Attenuates the magnitude of the forward-path transfer function
near and above the zero-gain crossover frequency. This improves
the relative stability of the system, but generally slows it down.
3. Suppresses high frequency noise
4. Reduces steady-state error.
 PID Tuning

 There are many techniques for tuning PID

 Zeigler-Nichols Step Response

 Suitable for process control where time lag is significant

k

0.63k

L T
k
The above response can be approximated by G (s)  e  sL
1  sT
where k is static gain, T time constant, L time delay and a  k L
T
Based rule-of-thumb, for 0.1  L T  0.6 the controller gains can be appoximated by

Controller Kp Ti Td
PID 1/a 3L L/2
PI 0.9/a 2L
P 1.2/a
 Zeigler-Nichols Closed-Loop Method

 Using Routh-Hurwitz criteria, we can determined the margainally stable’s gain, K c and its period
of oscilation, T for a controller with a gain of K
 With rule-of-thumb of thumb

Controller Kp Ti Td

PID 0.5 K c 0.833T 0.125T

PI 0.45 K c 0.5T
P 0.6 K c
Example
Design a P, PI and PID controller for a plant with an open loop transfer function as follows

6
G ( s) 
( s  1)( s  2)( s  3)

With a gain of K connected in cascade to the system, its characteristic equaton becomes

D( s)  s 3  6s 2  11s  6(1  K )  0

Forming the Routh array

1 11
s3
2 6 6(1+K)
s
10  K
s1
1+K
s0
Thus

K c  10 and its frequency of osccilation  c  11 which gives T= 1.895 s

This gives

Controller
Kp Ti Td
P 5

PI 4.5 1.57 0.237

PID 6 0.947

Their step response are

 Effect of adding zero and pole

Example

Consider an open-loop transfer function

25
G(s) 
(s  5)(s 2  2s  5)
By introducing a zero and normalizing the response

( s  a) 25
G ( s) 
a ( s  5)(s 2  2s  5)

The step response for uncompensated, a=4 and a=8.

The effect of introducing zero

 Reduce rise time, peak time

 Increases overshoot
 As the zero approach the orgin its contribution is more significant
 Bandwidth increase
 Improve gain margin

For a compensator with a pole and normalizing the output

a 25
G(s) 
( s  a) ( s  5)( s 2  2s  5)

The step response for uncompensated, a=4 and a=8.

The effect of introducing pole
• Reduce oscillation, and as its more dominant the response becomes sluggish
 Design in time response

Lag compensator
Use for steady state improvement without affecting the transient response
Consider a lag compensator transfer function

Ge ( s ) 
 s  ze 
 s  pe 
where ze  pe
.

ze pe
Consider a plant
m

 (s  z i )
G p (s)  K i 1
n

 (s 
j 1
pj)

Dc gain without compensator

z1  z2  ....  zm
Ku  K
p1  p2  ....  pn
where z1 , z 2 , ....z m and p1 , p 2 , .... p n zeros and poles position from test point.

With compensator
ze z1  z2  .... zm
Kc  K 
pe p1  p2  .... pn

As ze  pe , thus K  K with that the steady state error is reduced.

c u
Lead compensator
 For transient response

s  zd
Gd 
s  pd
where pd  z d .

Uncompensated
  s  z1    s  z2   .....   s  zm  
   s  p1    s  p2   .....   s  pn    r

Compensated
 s  zd    s  pd     s  z1    s  z2   .....   s  zm  
   s  p1    s  p2   .....   s  pn    r

Thus
 c   s  zd    s  pd 
Example:
s  zd
Use a lead compensator Gd  so that the overall system has undamped frequency
s  pd
of 4 rad.s-1 and rate of decay 0.5 s-1, if z d  12

+ 20 K
( s  1  j )( s  1  j )
-

1
s  10

Open loop transfer function

20 K 1 20 K
Go ( s)  . 
( s  1  j )( s  1  j ) s  10 ( s  1  j )( s  1  j )( s  10)
Dominant poles

s1, 2   n  j n 1   2
1
  n  0.5 and n  4

that give a damping ratio of

0.5 0.5
   0.125
n 4

and

 d   n 1   2  4 1  0.125 2  3.97

Third pole is determined from

p3  0.5  0.5  12
 p3  11
 j s-plane

c j3.97
 j
c 
-11-10 -1 -0.5
 -j
c -j3.97
Hence, the dc gain

20K
1
( s  1  j )( s  1  j )( s  10) s 0.5 j 3.97

20K  (10  0.5) 2  3.97 2  (1  0.5) 2  (3.97  1) 2  (1  0.5) 2  (3.97  1) 2  154.9

Which give the closed loop transfer function

154.9
( s  11)(s  0.5  j3.97)( s  0.5  j3.97)
Location of zero and pole of the system together with the compensator.

j Satah s

s j 3.97
 j
 O  
p d -12 -10 -1 -0.5 -j

 j 3.97
Using the angle condition on the closed loop poles of  0.5  j 3.97
 3.97   3.97   3.97  1   3.97  (1) 
tan 1    tan 1    tan 1    tan 1    
  0.5  (12)    0.5  (10)    0.5  (1)    0.5  (1) 
 180 

where  the angle of pd from  0.5  j 3.97

19.05  22.68  80.44  84.26    180

  11 .67 

i.e. the angle contribution required by

 3.97 
tan 1    11 .67 
  0.5  p d 
p d  19.72

Thus the compensator transfer function

s  12
Gd 
s  19.72
 Design in frequency response

Basic concepts in the design criteria

 Increase of phase margin, the overshoot will be reduced

 Increase of bandwith, the response will be faster
 Increase of low frequency magnitude, the steady state will be reduced
Gain adjustment
Use to reduce the overshoot during transeint period by incresing the phase margin

dB

g
0
g GM  (rad.s-1)

Plot LM

o PM
-180
PM’ Plot 

Design Procedure
(i) With a chosen gain, obtain the Bode plot and the gain crossover frequency, g
(ii) Determine the required phase margin
(iii) Obtain the gain that is required for the new cossover frequency,  g '
This gain is the extension of the gain in (i).
Example:
Consider a type-1 system.
(a) Determine the uncompensated phase margin.
(b) If a P-compensator is cascaded to the system, determine the required gain so that the
o
phase margin is 30.

+ 2  10 4
Gc s ( s  10)( s  500)
-
Solution:

(a) By replacing s  j , the frequency response in corner frequency form is

2  10 4
G ( j ) 
 j  j 
10  500  ( j )  1  1
 10  500 
4

 j  j 
( j )  1  1
 10  500 
and
 
 
 4 
LM ( j )  20 log  
    2     2  and
      1     1  
   10   500  
    

( j )  90   tan 1 (0.1 )  tan 1 (0.002 )

4
LM ( j1)  20 log   12 dB and
1
 ( j1)  90 
(rad.s-1) 1 10 500

-20 -20 -20

Pole at origin
1
s
(dB/dekad)
0 -20 -20
Real pole
1
s  10
(dB/dec)

Real pole 0 0 -20

1
s  500
(dB/dec)
Total slope -20 -40 -60
(dB/dec)

and
  (rad.s-1) 1 50 100 5000

Real pole -45 -45 0 0

1
s  10
(deg/dec)
Real pole 0 -45 -45 0
1
s  500
(deg/dec)
Total slope 45 -90 -45 0
(deg/dec)

LM 2  LM 1   LM  log  2  log 1 
and
 2  1     log  2  log 1 
 1 10 500 10000
(rad.s-1)
Total slope -20 -40 -60 -60
 LM
(dB/dec)

LM (dB) 12 -8 -76 -154

LM 1  LM

LM 2
LM 2  LM 1   LM  log  2  log  1 
log  1 log  2
 2   1     log  2  log 1 
  (rad.s-1) 1 50 100 5000 10000
Total -45 -90 -45 0 0
slope
(deg/dec)

(deg) -90 -166 -193 -269 -269

>> w1=[1 10 500 10000];LM=[12 -8 -76 -154];

>> w2=[1 50 100 5000 10000];ph=[-90 -166 -193 -269 -269];
>> subplot(2,1,1);semilogx(w1,LM);
subplot(2,1,2);semilogx(w2,ph);

From the Bode plot,  g  4 rad .s and PM  60 for uncompensated system

1 
That gives the compensator gain as
50

0
Gm'
-50

-100

-150

-200
0 1 , 2 3 4
10 10 10 10 10

-50

-100

-150

-200

-250

-300
0 Wg 1 Wg' 2 3 4
10 10 10 10 10

For the new phase margin, PM '  30  g '  13 rad .s

 1
d an increase of gain
 25 
Gm' 25dB That gives the compensator gain as Gc  antilog   17.8
 20 
Lag compensator
The compensator can also be presented as
Ts  1
Gc ( s ) 
Ts  1
1
s
1 T

 s 1
T
where   1 . The pole is chosen to be closed with the origin, example
s  0.1
Gc ( s) 
s  0.01
 As the zero is far away from the pole, no obvious transient response will be affected
 Increased in the open-loop gain will improve the steady state error
Asymtote approximation of the Bode plot
 dB 1 T 1T
 (rad.s1)

10 T 0.1 T  (rad.s-1)

Example of actual plot for   2 and 5, T  0.1 ,

>> alpha1=2;T=0.1;alpha2=5
>> sys1=tf([1 1/T],[1 1/(alpha1*T)]);
>> sys2=tf([1 1/T],[1 1/(alpha2*T)]);
>> bode(sys1,sys2)
Gambarajah 6.47: Plot Bode bagi pemampas mengekor dengan penghampiran asimtot
Observation from the plot:

 At low frequency, the compensator will not have influenced on the open-loop
gain so we can incresed the gain for steady state improvement

Design Procedure
Prosedur rekabentuk adalah seperti berikut :

(i) Determine the gain K for the required steady state error and draw the Bode plot
o o
(ii) Obtained the required phase margin and add, compensator phase angle, =5 - 12
(iii) Compensator’s zero will be a decade below of the new gain crossover frequency of (ii).
(iv) Compensator’s pole is obtained by drawing a slope of -20dB/dec from the compensator’s zero
until the line touched the 0dB.
Example

If the new gain crossover frequency is,  g ' , and the LM at that frequency is LM  '  Y
g dB.

Compensator’s zero 1 T  0.1 g '

While, compensator’s pole is obtained by finding the frequency where the slope of -20dB/dec which
begin from the compensator’s zero cross the 0dB-axis.

dB
1 T 1T

0  (rad.s-1)
20dB/dekad
-Y

(iv) Adjust the gain K for the specified error coefficient

Example:
Design a lag compensator so that the ramp error coefficient is 10 ses-1, gain margin of at
least 20 dB and phase margin of at least 45o. Assume the compensator’s phase angle is 5o.

+ K
Gc
s ( s  5)( s  20)
-
2
s 1

Open loop transfer function

2 KGc
s( s  1)( s  5)( s  20)

Consider G c ( s )  1 and K=50, which give an open loop transfer function of

2  50
G ( j ) 
 j  j 
5  20  ( j )( j  1)  1  1
 5  20 
1

 j  j 
( j )( j  1)  1  1
 5  20 
  
 
 1 
LM ( j )  20 log  
and
 
   2   2 
 ( )  2  1     1     1  
  5   20  
   

( j )  90   tan 1 ( )  tan 1 (0.2 )  tan 1 (0.05 )

>> w1=[0.1 1 5 20 100 1000];LM=[20 0 -28 -64 -120 -200];

>> w1a=[0.1 1000];LMa=[0 0];
>> subplot(2,1,1);semilogx(w1,LM,w1a,LMa,'-');
>> w2=[0.1 0.5 2 10 50 200 1000];ph=[-90 -121 -176 -270 -333 -360 -360];
>> w2a=[0.1 1000];pha=[-180 -180];
>> subplot(2,1,2);semilogx(w2,ph,w2a,pha,'-');
From the Bode plot
 p  2.1 rad ..s 1
 g  1 rad ..s 1
PM  30
GM  13 dB
Given

2K
K r  had sG ( s)   10
s 0 100

Open loop transfer function without compensator

2K
G(s) 
s( s  1)( s  5)( s  20)
with initial value of
 2  500 
LM ' ( j 0.1)  20 log    40 dB
100 ( 0 .1) 

Thus shift 20dB upward from original plot .

New phase margim with =5o, PM’=45o+5o=50o. thus g ’=0.65 rad.s-1.

LM ( g )  20dB

Hence the compensator must provide –20.

 >> w1=[0.1 1 5 20 100 1000];LM=[40 20 -8 -44 -100 - 180];

>> w1a=[0.1 1000];LMa=[0 0];

>> subplot(2,1,1);semilogx(w1,LM,w1a,LMa,'-');title('PLOT
BODE');ylabel('Magnitud (dB)')
>> w2=[0.1 0.5 2 10 50 200 1000];ph=[-90 -121 -176 -270 -333
-360 -360];
>> w2a=[0.1 1000];pha=[-180 -180];At
>> w2b=[0.1 1000];phb=[-130 -130];’,
>>subplot(2,1,2);semilogx(w2,ph,w2a,pha,w2b,phb,'-');xlabel('Frek
uensi (rad/s)');ylabel('Sudut fasa (Darjah)')
At  g ’, 1 T  0.065
Hence pole of the compensator

dB
0.0065 0.065
0  (rad.s-1)
-20dB/dekad
-20

1 1 T 0.065
Gain of the compensator    0 .1
 1T 0.65
s  0.06
The compensator Gc ( s)  0.1
s  0.0065
>> syms s;
>> den=expand((s+0.0065)*s*(s+5)*(s+20)*(s+1))
den = s^5+52013/2000*s^4+125169/1000*s^3+1613/16*s^2+13/20*s
>> num=expand(0.1*(s+0.065)*2)
num = 1/5*s+13/1000
>> num=[1/5 13/1000];den=[1 52013/2000 125169/1000 1613/16 13/20 0];
>> bode(num,den)
>> syms s;
>> den=expand((s+0.065) *s*(s+5)*(s+20)*(s+1))
den = s^5+5213/200*s^4+12669/100*s^3+865/8*s^2+13/2*s
>> num=expand(0.1*(s+0.65)*2)
num = 1/5*s+13/100
>> num=[1/5 13/100];den=[1 5213/200 12669/100 865/8 13/2 0];
>> bode(num,den)
>> num=[1/5 13/100];den=[1 5213/200 12669/100 865/8 13/2 0];
bode(num,den)
sys=tf(num,den);
[GM,PM,Wg,Wp] = margin(sys)
GM_dB = 20*log10(Gm)
GM = 693.1782
PM = 73.8732
Wg = 1.0751
Wp = 0.0192
GM_dB = 56.8169
System type –depends on input
"Weakness of attitude becomes
weakness of character."

"Anyone who has never made a

mistake has never tried anything
new."