0% found this document useful (0 votes)
19 views

Introduction To Automatic Control Week8-Slides

This document discusses lag compensators and their frequency response characteristics. It introduces lag compensators as having the form C(s) = K/(1+s/ω1)/(1+s/ω2) with ω2 < ω1, making them similar to proportional-integral controllers. The bode plot of a lag compensator shows a low frequency gain boost that saturates at ω2 rather than continuing to increase, and a low frequency phase lag that is attenuated at ω2 rather than continuing down to -90 degrees. Lag compensators can provide many of the same benefits as PI control in regulating steady-state error while avoiding some of the limitations.

Uploaded by

jonah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
19 views

Introduction To Automatic Control Week8-Slides

This document discusses lag compensators and their frequency response characteristics. It introduces lag compensators as having the form C(s) = K/(1+s/ω1)/(1+s/ω2) with ω2 < ω1, making them similar to proportional-integral controllers. The bode plot of a lag compensator shows a low frequency gain boost that saturates at ω2 rather than continuing to increase, and a low frequency phase lag that is attenuated at ω2 rather than continuing down to -90 degrees. Lag compensators can provide many of the same benefits as PI control in regulating steady-state error while avoiding some of the limitations.

Uploaded by

jonah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 37

Introduction to Automatic

Control
8 - Compensator Design
Dr Sebastian East
Contents

1. Overview

2. Lag Compensators
Frequency Domain
Root Locus

3. Lead Compensators
Frequency Domain
Root Locus

4. Lead-Lag Compensators

2 / 37
Intro
• Last week we developed a control theoretic understanding of
the principles of PID controller design.

• There is nothing (theoretically) ‘special’ about PID control


I The same theory can be used to propose, analyse, and design any
LTI controller.

• This week we will investigate compensators


“ ”
1
!1
s +1
C(s) = K “ ”
1
!2
s + 1

and their benefits/limitations relative to PI, PD, and PID control.

• As with last week, these slides will generally be conceptual.

3 / 37
Why ‘Compensators’?

• Why are these form of controllers called ‘compensators’ as


opposed to ‘controllers’?

• No good answer
I Just think of them as a particular form of LTI control structure.

R(s) + Y (s)
C(s) G(s)

• Beware - an internet search for this question will yield some


absolute nonsense.

4 / 37
Contents

1. Overview

2. Lag Compensators
Frequency Domain
Root Locus

3. Lead Compensators
Frequency Domain
Root Locus

4. Lead-Lag Compensators

5 / 37
Compensator Structure

• The general compensator structure is


“ ”
1
!1
s + 1
C(s) = K “ ”
1
!2
s +1

• The type of compensator is determined by the relative values of


!1 and !2 :
I !2 < !1 : Lag compensator.
I !1 < !2 : Lead compensator.

• This talk will focus on Lag compensators


I Spoiler: integral-type controller.

6 / 37
Proportional Integral Recap
• Recall the bode plot of the PI controller:
C(s) = kp + ksi = K (s+!
s
1)

20 log10 |C(i!)|
Low frequency magnitude
PI
boost (0 steady state error for
20 stable systems)
20 log10 K
!
!1 !1
10

90 \C(i!) Accompanying -90◦ phase lag


45
!
!1 !1 10!2
−45 10
−90 PI

7 / 37
Lag Compensator Frequency Response
• Lag compensator given by
“ ”
1
!1
s +1
C(s) = K “ ”
1
!2
s +1

with !2 < !1 .

• The characteristics of compensators are most easily interpreted


in frequency domain

• Compensator consists of three components:


„ «
1 1
K ד
C(s) = |{z} ”× s +1
1
s +1 !1
gain !2 | {z }
| {z } First Oder Lead
First Oder Lag

8 / 37
First Order Lag/Lead Bode Plot
• Bode plots of first order lag and first order lead:

20 log10 |C(i!)|
( !11 s + 1)
20
!
!2 !1
−20
−40 1
( !1 s+1)
2

90 \C(i!)
45 ( !11 s + 1) !
1
!2 ( !1
!1
−45 2
s+1)

−90

9 / 37
Combined Bode Plot
• The two previous Bode plots combine to:

20 log10 |C(i!)|
( !11 s + 1)
20
!
( !1 s+1)
1
!2 !1
−20 ( !1 s+1)
2

−40 1
( !1 s+1)
2

90 \C(i!)
45 ( !11 s + 1) !
1
( !1 s+1) !2 ( !1
!1
−45 1
2
s+1)
( !1 s+1)
−90 2

10 / 37
Fixed Gain
• Relation to PI control (discussed later) made easier by
introducing fixed gain !!12 :

40 20 log10 |C(i!)|
( !11 s + 1)
20
!1 ( !1 s+1) !
1
!2 (1!1 s+1)
( ! 2s+1)
1
!2 !1
−20 ( !1 s+1)
2

−40 1
( !1 s+1)
2

90 \C(i!)
45 ( !11 s + 1) !
1
1 !2 !1
−45 !1 ( !1 s+1) ( !1
2
s+1)
!2 ( !1 s+1)
−90 2

11 / 37
Chosen Gain
• Arbitrary gain boost then provided by chosen value of K̂ (so that
K = K̂ !!21 :

20 log10 |C(i!)|
40
( 1
s+1) ( !11 s + 1)
20 log10 K̂ K̂ !!21 ( !11 s+1)
!2 !
( !1 s+1) !2 !1
-20 ( !1
1
s+1)
-40
2
1
( !1 s+1)
2

90 \C(i!)
45 ( !11 s + 1) !
1
1 !2 !1
−45 K̂ !1 ( !11 s+1) ( !1
2
s+1)
!2 ( ! s+1)
−90 2

12 / 37
Comparison with PI
• Lag Compensator is similar to PI controller, except two
differences:

20 log10 |C(i!)|

PI Low frequency magnitude


40 boost saturates at !2
Lag Comp.
20 log10 K̂
!
!2 !1

90 \C(i!) Low frequency phase lag is


45 attenuated at !2
!
! !1
−45 Lag Comp. 2
−90
PI

13 / 37
Effect of !2
• Design process for Lag Compensator is the same as for PI, with
the exception of also choosing !2
I 0 s.s. error no longer guaranteed.

14 / 37
Lag Compensator vs PI

• Why would one want to use a Lag Compensator instead of PI?

• General (but not very helpful) answer:

When the characteristics of the controller help you achieve


given design requirements.

• Possible case:
I PI introduces phase lag, and increases gain at frequencies ! < !1
I Both of these changes can reduce margins, so Lag Comp. could
be preferable.
I But, you could just make !1 lower...

• Internet is again riddled with nonsense.

15 / 37
A Possible Use Case

• Recall that for PI, |C(i!)| → ∞ as ! → 0

• In practice this is generally impossible


I Output of controller circuit will saturate.
I Control input will saturate.

• Therefore, |C(i!)| will saturate anyway


I But, this isn’t reflected in either the Bode plot or root locus

• A lag compensator allows you to design you controller with the


knowledge of where the gain will saturate.

16 / 37
Contents

1. Overview

2. Lag Compensators
Frequency Domain
Root Locus

3. Lead Compensators
Frequency Domain
Root Locus

4. Lead-Lag Compensators

17 / 37
Root Locus Interpretation
• Recall that PI controller equivalent to

1 (s − –1 )
C(s) = ki + kp ⇔ K
s s
• Recall that the root locus process of designing a PI controller is
1. Place open-loop pole at 0
2. Choose open-loop zero at s = –1
3. Choose gain K that places dominant poles in desired region.

• For lag compensator, we can now choose open loop pole on


real axis. “ ”
1
!1
s + 1 (s − –1 )
C(s) = K “ ” ⇔L
1
s +1 (s − –2 )
!2

• Generally, we want to place –2 as close as possible to zero


I Limited by the performance of hardware.

18 / 37
Conclusion

• Lag compensator is an integral type controller.


I Low frequency magnitude boost reduces steady-state errors.
I 0 steady state error no longer guaranteed.
I Lag compensator explicitly accounts for input saturation (can be
used as part of an anti integral-windup strategy)

• Root locus now requires the choice of open loop pole

• Next we will look at the Lead compensator


I Spoiler: it’s a saturated PD controller!

19 / 37
Contents

1. Overview

2. Lag Compensators
Frequency Domain
Root Locus

3. Lead Compensators
Frequency Domain
Root Locus

4. Lead-Lag Compensators

20 / 37
Lead Compensation

• Recall that the general compensator structure is


“ ”
1
!1
s + 1
C(s) = K “ ”
1
!2
s +1

• We will now take a look at Lead Compensators:


I !2 < !1 : Lag compensator.
I !1 < !2 : Lead compensator.

• Principles are generally the same as for lag compensators -


won’t go through in as much detail.

21 / 37
PD Controller Recap
• Recall the bode plot of the PD controller:
C(s) = kp + kd s = K( !11 s + 1)

20 log10 |C(i!)|
High frequency
20 magnitude boost
PD
20 log10 K
!
!1 10!1

90 \C(i!)
45 PD !
!1 10!1
−45
+90◦ phase lead
−90

22 / 37
Lead Compensator Lead/Lag
• When !1 < !2 , the first order lead occurs before the first order
lag:

20 log10 |C(i!)| ( !11 s + 1)


20
!
!1 !2
−20 1
( !1 s+1)
−40 2

90 \C(i!)
45 ( !11 s + 1)
!
1
!1 !2( !1 s+1)
−45 2

−90

23 / 37
Combined Response
• Following the same steps as before, this results in the same
response, reflected in vertical axis:

20 log(101 |C(i!)|
s+1)
40 K ( !11 s+1) ( !11 s + 1)
!2
20 log10 K
!
( !1 s+1) !1 !2
-20 1
( !1 s+1) 1
-40
2 ( !1 s+1)
2

90 \C(i!)
45 ( !11 s + 1)
!
( !1 s+1) 1
K 1 !1 !2( !1 s+1)
−45 ( !1 s+1) 2
2
−90

24 / 37
Comparison with PD
• Lead compensator is similar to PD controller, except two
differences:

20 log10 |C(i!)|
High frequency magnitude PD
boost saturates at !2
40 Lead Comp.
20 log10 K
!
!1 !2

90 \C(i!) PD
45
!
!1 !2 Lead Comp.
−45
High frequency phase lead is
−90
attenuated at !2

25 / 37
Lead Compensator vs PD

• Why would one want to use a Lead Compensator instead of PD?

• Same (and still not very helpful) answer as before:

When the characteristics of the controller help you achieve


given design requirements.

• Possible case:
I Previously discussed how derivative gain amplifies noise
I This effect is automatically reduced by saturating the PD amplitude
ratio at a given value.

26 / 37
Alternative Interpretation

• Recall that a method for noise attenuation in derivative control is


to prefilter the derivative action:

1
Error Signal fi s+1 kd s Derivative Action

• Lead compensator can be interpreted as a prefiltered PD


controller:

1 ”
“ ”

Error Signal 1
!2 s+1 K 1
!1
s +1 Lead Comp.

27 / 37
Example
• Consider the nominal system G(s) = 0.1s+1
s 2 +2s+1

20 log10 |G(i!)| !
10−1 100 101 102 103 104
−40

−80
G(s)

\G(i!) !
10−1 100 101 102 103 104
−45
−90 G(s)
−135

28 / 37
PD Control
• Consider PD control C(s) = 0.01s + 1:

20 log10 |G(i!)| !
10−1 100 101 102 103 104
−40
PD
−80
G(s)

\G(i!) PD !
−1 0 1 2 3
10 10 10 10 10 104
−45
−90 G(s)
−135

29 / 37
Lead Compensator
• Consider lead compensator C(s) = 0.01s+1
0.001s+1
:

20 log10 |G(i!)| !
10−1 100 101 102 103 104
−40
PD
−80 Lead
G(s)

\G(i!) PD !
10−1 100 101 102 103 104
−45 Lead
−90 G(s)
−135

30 / 37
Contents

1. Overview

2. Lag Compensators
Frequency Domain
Root Locus

3. Lead Compensators
Frequency Domain
Root Locus

4. Lead-Lag Compensators

31 / 37
Root Locus Interpretation
• Same as lag: PD design but can now choose pole on real axis.
• Pole and zero generally chosen further to left than system
dynamics, and !2 > !1
• Zero ‘pulls’ the root locus, pole ‘pushes it’
• Order of system stays the same, so do number of asymptotes.

32 / 37
Conclusion

• Lead compensator is a derivative type controller.


I High frequency boost helps improve damping (rejects high
frequency disturbances).
I Can be interpreted as a pre-filtered PD controller

• Next we will look at the Lead-lag compensator


I Spoiler: it’s a saturated PID controller!

33 / 37
Contents

1. Overview

2. Lag Compensators
Frequency Domain
Root Locus

3. Lead Compensators
Frequency Domain
Root Locus

4. Lead-Lag Compensators

34 / 37
Lead-lag Compensation
• Final piece in the puzzle is the lead-lag compensator

• Nothing more than a cascade of a lead compensator and a lag


compensator (or vice versa):
“ ” “ ”
1 1
s+1 s+1
Error Signal Lead-lag Comp.
! !
K1 “ 1
1
” K2 “ 3
1

!2 s+1 !4 s+1

• Equivalent to
“ ”“ ”
1 1
!1
s +1 !3
s +1
C(s) = K “ ”“ ”
1 1
!2
s +1 !4
s +1

• Five parameters to choose: K, !1 , !2 , !3 , and !4 .

35 / 37
Comparison with PID
• Can be interpreted as a PID controller with saturation on both
amplitude boosts and attenuation of phase lead/lag:

20 log10 |G(i!)|

PID

Lead-lag
!

90 \G(i!)
Lead-lag !

−90 PID

36 / 37
Conclusion
• We have introduced compensators as an alternative form of
controller

• Lag compensator is equivalent to PI with the amplitude gain


saturated, and phase lag attenuated.

• Lead compensator is equivalent to PD with the amplitude gain


saturated, and phase lead attenuated.

• Lead-lag compensator is equivalent to PID with both.

− Much more common for the limitations of PID to be addressed


with alternative methods.
• Next week: state-space - the basis of all modern control
methods
I Optimal control (and estimation)
I Robust control
I Model predictive control
I ...
37 / 37

You might also like