Lecture 10:
Compensator
Design by
Root Locus
Instructor
Dr Samina Jamil
1
2
Lecture 10: Design
by Root Locus
Lecture Overview
•Compensator
•Dominant poles
•PI compensators
•Lag compensation
•Lead compensation
•PID controller
All pole on the radial line (a 3
line drawn from origin) have
same ζ and hence %OS .
Lecture 10: Design
by Root Locus
Compensator (Dynamic Controller) poles/zeros) to it to change the shape of root
•Let a system to be controlled has pole at A. locus.
•This additional gain with poles/zeros is
•Its %OS is fine, but rise time is undesirable.
called compensator.
•If pole location is changed to B, it gives
•Through compensator, improvement
desired rise time and %OS remains the same.
can be achieved in transient response (by
•Since changing the K will change pole adding differentiator) and the ess can be
location only on this root locus, we must find a reduced (by adding an integrator) etc.
new root locus that may pass through B without
changing damping coefficient.
•We can not change the system (plant), but
we can add an additional complex gain (more
4
Lecture 10: Design
by Root Locus
Dominant Poles •Output of actual system should be like
•The poles closer to the jω axis are called approximated system.
dominant poles. •When a higher order system is
approximated satisfactorily to 2nd order
Dominant Poles Approximation system, all techniques for 2nd order systems
•A higher order system is approximated by can be applied for its analysis and design.
2nd order system if it has some dominant poles •If a higher order system has zeros as
and other poles very far from dominant poles. well, and if these zeros have poles close to
•Poles at least 5 times farther from dominant them, they may be considered cancelled.
poles can be omitted for approximation.
5
Lecture 10: Design
by Root Locus
Dominant Pole Approximation that the approximation is not good.
Example 4.8 (Nise) •Another system, T2(s) has same
•T3(s) is a 3rd order system with poles -3, dominant poles but 3 rd
pole is very far.
-24.53j. •c1(t), the step response of T1(s) is similar
•To approximate it to a 2nd order system T1(s), to c2(t) showing that the approximation is
pole other than dominant is ignored. good.
•To check similarity, compare step response
of the two systems.
•c3(t) is much different from c1(t) showing
6
Lecture 10: Design
by Root Locus
ζ , ωd and ωn on Root Locus
• The damping ratio ζ and undamped natural
frequency ωn can be measured from root locus if a Example 13.40 (Schaum) Determine pole
radial line is drawn on a close-loop pole (on root location on root locus which gives a damping
locus) making an angle φ with –ve real axis. ratio of 0.55?
•The relation is •Since ζ=cos φ
• The length of line from origin to root locus 0.55=cos φ
pole = ωn φ=cos-1 0.55=55.6o
•Then we can find all specifications of system: • Required pole location is s=-0.7+1j at
K=7.
7
Lecture 10:
Compensator Design
Compensation Configurations
Cascaded compensation
•Compensator is added in series of plant.
•It operates on error signal.
•Will discuss only Cascaded compensation.
Feedback compensation
•Compensator is added in feedback loop.
•Control signal (O/P of compensator) is
compared with reference to generate the error.
8
Lecture 10: Cascaded
Compensator Design
Ideal Integral/PI Compensation
•The ess can be removed by increasing system
type; by adding an integrator (pole at origin).
•This type of compensator is implemented
using active circuits (op-amp etc.).
•Ideal integrator has a proportional gain as
well, hence also called Proportional + Integral,
(PI) compensator.
9
Lecture 10: Cascaded
Compensator Design
Ideal Integral/PI Compensation
•Let this uncompensated system has •pole and ess is removed as system is type 1.
desirable transient response but has some steady
state error as well.
•To remove ess, add an integrator. This
changes the root locus and now A is no more a
close-loop pole (not on root locus). But we want
this close loop pole A on same location.
•Add a zero close to origin, root locus returns
to its required shape as zero almost cancels the
10
Lecture 10: Cascaded
Compensator Design
Ideal Integral/PI Compensation •The complex pole pair is at -0.694j3.926
Example 9.1 (Nise) Design the system with •Using angle and magnitude criteria,
damping ratio=0.174 and zero position ess? K=164.6 and the 3rd pole is at -11.6.
Uncompensated system
•System is type 1 so has finite position ess.
•The root locus is:
•Using radial line at ζ=cos φ
0.174=cos φ
φ=cos-1 0.174=79.98o from –ve real axis.
11
Lecture 10: Cascaded
Compensator Design
Ideal Integral/PI Compensation
Example 9.1 (Nise) Design the system with
damping ratio=0.174 and zero position ess?
Uncompensated system
•The position error is calculated as:
•The steady state response settles at 0.892.
>> G=tf(164.6,conv([1 3 2],[1 10]))
>> G1=feedback(G,1)
>> step(G1)
12
Lecture 10: Cascaded
Compensator Design
Ideal Integral/PI Compensation increases the system order from 0 to 1.
Example 9.1 (Nise) Design the system with
damping ratio=0.174 and zero position ess?
PI compensation
•Adding a pole at origin and a zero at -0.1.
•The root locus for the compensated system:
•Using radial line for same ζ :
•K=158.2 and integrator pole is at -0.0902
which effectively neutralizes zero at -0.1 but
13
Lecture 10: Cascaded
Compensator Design
Ideal Integral/PI Compensation shown:
Example 9.1 (Nise) Design the system with •The transient response is same and ess is
damping ratio=0.174 and zero position ess? zero after compensation.
PI compensation
>> G2=tf([158.2 15.82],conv([1 3 2],[1 10 0]))
>> G3=feedback(G2,1)
>>eig(G3)
>>step(G1,G3)
•The step responses of uncompensated
(blue) and compensated (red) schemes are
14
Lecture 10: Cascaded
Compensator Design
Implementation of PI Compensation
•A possible method is as shown:
•K2 is used to achieve desired value of zero.
15
Lecture 10: Cascaded
Compensator Design
Physical PI Compensator Circuit
General active circuits
•The gain of amplifier is Gc(s)= •Then Gc(s)=
Active PI compensator
•Can be implemented only through active
•Comparing with
circuit.
•Let Z1=
•K1=-R2/R1 and K2/K1=1/R2C
•Z2=
16
Lecture 10: Cascaded
Compensator Design
Lag Compensation •They are easy to implement and cost
•Adds a pole near but not on the origin. less.
•Can be implemented both with passive or
active circuits.
•Lag compensators do not increase system
type.
•Nevertheless, they reduce ess .
•For sinusoidal inputs, a lag compensator
produces a phase lag at the output, hence they
are called lag compensators.
17
Similar theory holds true for type 0 and type 2 etc.
Lecture 10: Cascaded
Compensator Design
Lag Compensation response change).
•An uncompensated type 1 system is:
•The static velocity error constant is:
•Using a lag compensator with pc closer to •zc >zp reduces ess by increasing Kv.
origin than zc.
•Keep zc and pc close to origin (to reduce ess)
and close to each other (to nullify transient
18
Lecture 10: Cascaded
Compensator Design
Lag Compensation
Example 9.2 (Nise) Design lag compensator for •The step response of uncompensated
type 0 system in Example 9.1 with ξ=0.174 and and the compensated systems is:
improvement of 10 times in ess?
•From Example 9.1, the K=164.6, ess=0.108.
•The improved ess should be 0.0108, so
•Let pc= - 0.01 then zc= - 0.111
19
Lecture 10: Cascaded
Compensator Design
Lag Compensation
Physical lag compensator
•Can be implemented with active or passive
circuits. Passive circuits
Active circuits
•A possible method is as shown: •The gain Gc(s)=
•The gain Gc(s)=
•zc is always farther from origin than pc.
20
Lecture 10: Cascaded
Compensator Design
Derivative and Lead Compensation
•Integral and lag compensation improve
steady state error.
•Derivative, lead and proportional derivative
(PD) compensators speed up transient response.
21
Lecture 10: Cascaded
Compensator Design
PD compensation •Z2=
•One configuration of PD compensator is: •Then Gc(s)=
Physical circuit Can be implemented only
through op-amp based circuit
•Let Z1=
22
Lecture 10: Cascaded
Compensator Design
Lead compensation
•The pc is farther from origin than zc.
Active circuit
•The gain Gc(s)=
Passive circuit
•The gain Gc(s)=
23
Lecture 10: Cascaded
Compensator Design
Lag-Lead Compensation
•The scheme is used to reduce ess and
transient response simultaneously.
Active circuit
•The gain is the product of lag and lead gains
Passive circuit
• Gc(s)=
24
Lecture 10:
PID Design
PID Compensation
•PID compensator is as shown •Then Gc(s)=
R2 C2
C1
Active circuit
•Can be implemented only through active
circuit:
•Let Z1= and Z2=
R1
25
Lecture 10:
PID Design
PID Compensation
•Example 9.9 (Nise) PID compensator
designed through root locus method is as shown.
Construct the circuit to realize this PID?
•Rewriting
•Comparing with
26
Lecture 10:
PID Design
PID compensation time and Td the derivative time.
Ziegler-Nichols rules •There are two different methods of tuning a
PID using Ziegler-Nichols rule.
•These are the set of rules to design a PID.
•Works even if the plant model is not known.
•This set of rules gives an educated guess for
gains Kp, Ti and Td to start with.
•The gains are tuned based on plant’s step
response.
•Kp is called proportional gain, Ti the integral
27
Lecture 10:
PID Design
Ziegler-Nichols rules
Method 1
•Step response of actual/simulated plant is •K is the steady-state value of output.
obtained. •K, the line intercept gives time constant
•If plant does not have pure integrator or T.
complex dominant poles, then response typically •The first order approximation of the
looks like S-shaped curve. plant is:
•Draw a tangent on curve at point of
inflection.
•Time axis intercept determines L, the delay.
28
Lecture 10:
PID Design
Ziegler-Nichols rules •PID controller has a double zero at -1/L
and a pole at origin.
Method 1
•For PID controller the values are selected
from table:
•Hence
29
Lecture 10:
PID Design
Ziegler-Nichols rules are used to define parameters of PID as:
Method 2
•First Ti= ꚙ and Td = 0 is set.
•Hence PID operates only with Kp’ .
•Value of Kp’ is increased from zero to Kcr
(critical gain) until output exhibits sustained
oscillations. If no oscillations result for any value
of Kp’ , this method is not valid.
•The Kcr and corresponding time period Pcr
30
Lecture 10:
PID Design
Ziegler-Nichols rules
Method 2
•Hence the PID controller obtained is:
•It has a double zero at -4/Pcr and a pole at
origin.
•This method can be applied using root locus
plot as well.
•Pcr can be found from frequency of jω axis
crossing. Kcr is the corresponding gain.
31
Lecture 10:
PID Design 1
Ziegler-Nichols rules
Example 8.1 (Ogata) Design a PID to control the
system so that Tr<1 sec and %OS=16%?
•The step response without controller is
•It has large rise time
32
Since plant has pure integrator,
method 2 will be used.
Lecture 10:
PID Design Kp ’
Ziegler-Nichols rules
Example 8.1 (Ogata) Design a PID to control the •Same result can be obtained from root locus
system so that Tr<1 sec and %OS=16%?
•Using Method 2, initially the PID is Kp’ only.
•Close loop TF is:
•Char. equation is:
•Kp=Kcr is found by Routh-Hurwitz
•Kp=30 makes jω crossing for the root locus.
33
Lecture 10:
PID Design
Ziegler-Nichols rules
Example 8.1 (Ogata) Design a PID to control the
system so that Tr<1 sec and OS=16%?
•Auxiliary equation becomes s2+30=0
•Evaluating at jω axis
•Time period of oscillation is:
•Using table
34
Lecture 10:
PID Design
Ziegler-Nichols rules
Example 8.1 (Ogata) Design a PID to control the •PID controller is
system so that Tr<1 sec and OS=16%?
•The gains are
35
Lecture 10:
PID Design
Ziegler-Nichols rules
Example 8.1 (Ogata) Design a PID to control the
system so that Tr<1 sec and OS=16%?
•Checking the step response:
•The step response has required Tr.
•Large %OS should be reduced.
•Fine tuning of gain is required.
36
Lecture 10:
PID Design
Ziegler-Nichols rules
Example 8.1 (Ogata) Design a PID to control the
system so that Tr<1 sec and OS=16%?
•Using MATLAB/Simulink construct the the
control system.
•Double click on PID block, then click on Tune
button. PID tuner app opens.
•Adjust responses through slider, then
update the model.
37
Lecture 10:
PID Design
Ziegler-Nichols rules
Example 8.1 (Ogata) Design a PID to control the
system so that Tr<1 sec and OS=16%?
•The gains are updated in model
•The step response is
•Ziegler-Nichols provides with a guess to
start with.
