2014 Lec 16 Loop Shaping

Control Systems
Control Design: Loop shaping

L. Lanari
Sunday, November 30, 2014
Outline
specifications
open-loop shaping principle
lead and lag controllers
4 basic situations
sketch of PID controllers
Lanari: CS - Loop shaping

Specifications (closed-loop)
Static
desired behavior w.r.t. order k inputs in terms of a maximum allowed absolute error
desired attenuation level w.r.t. constant disturbances acting on the forward loop
tracking of a sinusoidal reference
attenuation of sinusoidal disturbances and measurement noise
Dynamic
location of eigenvalues/poles in the complex plane

time domain specifications on the step response (mainly on the reference to output behavior)
frequency domain specifications through the resonance peak and the bandwidth (mainly on the
reference to output behavior)
Stability
location of eigenvalues/poles in the complex plane

robustness in terms of gain and phase margin

Specifications
CLOSED-LOOP System
Static
equivalent to
OPEN-LOOP System
presence of a sufficient number of poles in 0
and/or at the reference/disturbance angular
frequency sufficiently high gain (in absolute
value)
Necessary conditions require the following
structure in the controller
for sinusoidal
1
Kc
reference or
2
2
h
s +
s
disturbance
Dynamic
bandwidth B3 (and rise time tr)
!c crossover frequency
resonant peak Mr (and overshoot Mp)
PM phase margin
taken care by
R(s)
Kc
C(s) = h R(s)
s

d1
d2
+ +
C (s)
r
m+
Kc
sh
R (s)
P (s)
+
+
necessary part of the controller
C (s)
controller
Kc
sh
P (s)
R (s)
to be still chosen in order to satisfy the dynamic specifications
P (s)
extended plant
plant
Loop function L (s)
R (s)
dynamic
specifications

Kc
sh
P (s)
static
specifications
5
Open Loop Shaping

Basic idea: being the plant P (s) fixed, choose the controller C (s) to shape the loop function
frequency response L (j!)
L (j!)
C (s)
magdB
such that
with some desired

characteristics
phase
PM
C (s) = C1 (s)R (s)

C1 (s)
static specs
Kc
sh
R (s)
dynamic performance specs
static specs

hyp: the necessary part of the controller C (s) has been determined so we have the
extended plant
Kc
P (s) = C1 (s)P (s) = h P (s)

s
we need to determine C (s) so to satisfy also the dynamics specifications and ensure stability
specifications
!c* desired crossover frequency at which we want to have a phase margin of at least PM *
from the extended plant frequency response we need to check which action is necessary
both in terms of magnitude and phase by comparing the actual value of the magnitude and
phase at the future crossover frequency !c*
Amplification: we need to increase the magnitude at some frequency

Magnitude
Attenuation: we need to decrease the magnitude at some frequency
Lead: we need to increase the phase at some frequency

Phase
Lag: phase can be decreased at some frequency if necessary
we want P M P M
so if we have extra phase we can keep it

7
P (j)
attenuation needed
!c*
2 possible actions on the

magnitude at some frequency
|P (j)|dB
amplification needed
|P (j)|dB
maximum lag we can tolerate

lead (phase increase) needed
PM *
-
2 possible actions on the

phase at some frequency
P (j)

Remember that the static specifications have already been met (if we ensure stability) and
therefore we do not want to alter this first step.
Therefore, in general, we are not going to use a
zero in s = 0 to obtain a phase lead

pole in s = 0 to attenuate at some frequency
a gain smaller than 1 in magnitude to attenuate if we have a constraint on the loop gain from
the static requirements (while we may use a gain greater than 1 to amplify)
elementary functions that can provide these magnitude and phase contributions
lead
compensator
1 + a s
Ra (s) =
a s
1+ m
a
lag
compensator
i s
1+ m
i
Ri (s) =
1 + i s
a > 0
ma > 1
i > 0
mi > 1
both have unit gain so no magnitude change in ! = 0

Lead compensator
Magnitude (dB)
1 + a s
Ra (s) =
a s
1+ m
a
Phase (deg)
pole in
ma
a
zero
10
0
10
20
1
a
ma
a
pole
zero
45
0
45
pole
90

ma > 1
20
90
ma = 10
a > 0
zero in
1
a
10
Lag compensator
Magnitude (dB)
i s
1+ m
i
Ri (s) =
1 + i s
i > 0
mi > 1
20
10
1
i
0
10
20
pole
zero in
mi
i
pole in
1
i
zero
mi
i
Phase (deg)
90
mi = 10
0
45
90

zero
45
pole
11
Choice of R (s)
we assume C1 (s) (static specs) has already been chosen. Therefore we need to
by evaluating the actual values of the extended plant magnitude and phase at the desired
crossover frequency !c*, understand what action needs to be undertaken:
amplification or attenuation at !c*

phase lead or maximum allowed phase lag at !c*
choose the elementary function(s) Ra(s) and/or Ri(s) (since multiple actions can be
combined) needed
choose ma (or mi) and the normalized frequency !

deciding to obtain the desired action at !c* choose a (or i) so that
!c*a = !
frequency
chosen
at which we
want to obtain
the desired action

(or !c*i = !)
we solve
in a
12
Universal diagrams
25
16
1 + a s
Ra (s) =
a s
1+ m
a
ma > 1
12
10
20
8
magnitude (dB)
a > 0
14
15
4
3
10
2
5
0 1
10
for different values of ma
10
10
normalized frequency
10
70
16
60
50
phase (deg)
for Ri(s) just change sign

to the ordinates
14
12
10
8
6
40
4
30
3
20
2
10
0 1
10

10
10
10
13
Case I
40
20
30
R a (j)
0
20
20
10
Magnitude (dB)
Magnitude (dB)
40
40
60
2
10
10
10
frequency (rad/s)
10
Phase (deg)
50
0
10
20
L(j )
30
90
40
PM *
180
200
270
2
10
specs
10
10
frequency (rad/s)
!c* = !c
60
2
10
10
30
actions needed:
10
frequency (rad/s)
10
frequency which coincides with the desired one

phase: increase the phase, in this case of exactly
PM *
90
L(j )
180
200
P (j)
270
2
10
R a (j)
Phase (deg)
magnitude: as small amplification as possible in

order not to move the current crossover
10
PM PM *
50
P (j)
50
10
10
frequency (rad/s)
10
14
example we need a phase lead of 25 with the smallest amplification possible

70
16
60
phase (deg)
50
25

40
4
30
20
2
10
0 1
10
we have chosen
10
10
10
0.5 is the smallest (for

these plots) normalized
frequency at which we
obtain the desired phase
lead (for ma = 16)
25
16
14
12
10
20
8
magnitude (dB)
with the amplification of

1 dB at !c* = 0.3 rad/s
we choose a as
a = 0.5/0.3
desired
phase lead
ma = 16
! = 0.5
therefore to obtain
this lead of 25 together
several choices of ma are

possible (at different
normalized frequencies),
we need to find one
which is compatible with
the magnitude
requirement
14
12
10
15
4
3
10
smallest amplification
obtainable (around 1 dB)
2
5
0 1
10
1 dB
0
10
10
10
15
Case II
40
|P (jc )|dB
20
0
20
P (j)
L(j )
20
10
40
60
4
10
10
10
10
frequency (rad/s)
10
10
50
Phase (deg)
30
Magnitude (dB)
Magnitude (dB)
40
R i (j )
20
40
PM *
180
200
specs
10
30
90
270
4
10
10
10
10
frequency (rad/s)
10
50
60
4
10
10
10
10
10
frequency (rad/s)
10
10
PM PM *
!c*
50
30
actions needed:
magnitude: attenuation of |P (jc )|dB
phase: since
P (jc ) + > P M
Phase (deg)
R i (j )
90
P (j)
L(j )
180
we can tolerate at most a lag of

P (jc ) + P M

200
270
4
10
10
10
10
frequency (rad/s)
10
10
16
example we need an attenuation of 17 dB and can tolerate a maximum lag of 7

- 25
negative
values
required attenuation
16
12
- 20
10
magnitude (dB)
3
2
0 1
10
10
16
- 40
6
4
60 is the smallest (for

these plots) normalized
frequency at which we
obtain the desired
attenuation (for mi = 8)
- 20
14
12
10
- 50
- 30
maximum allowed lag
10
- 70
negative
values - 60
we choose i as
i = 60/0.1
10
phase (deg)
together with a lag

smaller than 7
at !c* = 0.1 rad/s
- 10
mi = 8
! = 60
therefore to obtain
this attenuation of 1 dB
several choices of mi are

we need to find one
which is compatible with
the magnitude
requirement
- 15
we have chosen
14
- 10
0 1
10
10
10
10
17
on the choice of the normalized frequency (case II)

basic consideration (see sensitivity functions): we usually want open-loop high gain at low
frequency, therefore anything that gives an excess of attenuation at low frequency should be
avoided if possible
case II with two alternative choices of the !
Ri1(s)
mi = 8
! = 70
Ri2(s)
mi = 8
! = 200
starts attenuation before strictly needed
80
0
70
R i1(j )
60
R i2(j )
P (j)
40
30
L 1 (j )
P (j)
L 2 (j )
20
Phase (deg)
Magnitude (dB)
50
10
90
L 1 (j )
L 2 (j )
0
180
R i1(j )
10
R i2(j )
200
20
30
4
10
10

10
frequency (rad/s)
10
10
10
10
10
frequency (rad/s)
10
10
both compensators solve the specifications

18
Case III
40
20
0
30
|P (jc )|dB
20
40
60
4
10
10
10
10
frequency (rad/s)
10
10
Phase (deg)
50
0
20
10
L(j )
20
R a (j) P (j)
40
PM *
180
200
specs
30
90
270
4
10
R a (j)
10
Magnitude (dB)
Magnitude (dB)
40
10
!c*
10
10
frequency (rad/s)
10
P (j)
50
60
2
10
10
10
10
frequency (rad/s)
10
PM PM *
50
30
actions needed:
magnitude: amplification of |P (jc )|dB
phase: since
P (jc ) + < P M
we need to obtain a phase lead of
P M P (jc ) +

Phase (deg)
R a (j)
90
180
200
L(j )
P (j)
270
2
10
10
10
frequency (rad/s)
10
19
example we need an amplification of 20 dB and a phase lead of at least 25

two choices
find a lead compensator that will give simultaneously the required amplification
and phase lead
usually requires a choice of the normalized frequency on the right-hand side

of the phase bell shape which gives poor robustness w.r.t. increases in the
crossover frequency since the phase of both the extended plant and the
compensator are decreasing at the chosen frequency
may be not easy to find
find a lead compensator that gives the required lead and gives some
amplification, integrate the required amplification with an additional gain Kc2
greater that 1.
This is usually possible since the static requirement (if any) asks for a gain
sufficiently high.

20
example we need an amplification of 20 dB and a phase lead of at least 25

70
Kc2Ra(s)
16
60
phase (deg)
50
25
8
6
40
4
30
3
20
desired
phase lead
2
10
we can choose
0 1
10
ma = 3
! = 0.9
therefore to obtain
this lead of 25 together
10
10
10
Kc2 /
[Kc2 ]dB = 20 - 2.5
25
16
14
12
10
20
8
magnitude (dB)
with the amplification of

2.5 dB at !c* = 0.2 rad/s
we choose a as
a = 0.9/0.2
several choices of ma are

we just keep the choice of
the normalized frequency
on the left of the bellshape
14
12
10
15
we can obtain an
amplification of 2.5 dB)
10
2
5
2.5 dB
0 1
10
10
10
10
21
about the robustness issue

case III revisited with two alternative solutions (the gains have been chosen appropriately)
70
16
60
Kc1 Ra1 (j)
14
12
phase (deg)
ma = 6
! = 0.8
Kc2 Ra2 (j)
10
50
8
6
40
ma = 6
! = 7.5
4
30
3
20
2
left-hand side of the bell
right-hand side of the bell
10
0 1
10
10
10
10
40
30
50
R a 1(j)
30
R a 2(j)
20
R a 2(j)
w.r.t. an increase of the

crossover frequency
R a 1(j)
Phase (deg)
Magnitude (dB)
10
0
10
20
L 2 (j )
90
L 2 (j )
30
180
L 1 (j )
40
10
10
frequency (rad/s)
L 1 (j )
P (j)
P (j)
50
60
2
10
200
10
270
2
10
10
10
frequency (rad/s)
10
solution 1 is more robust w.r.t. uncertainties in the !c

22
Case IV
40
|P (jc )|dB
20
0
20
20
R a (j)
10
40
60
2
10
10
10
frequency (rad/s)
10
50
Phase (deg)
30
Magnitude (dB)
Magnitude (dB)
40
10
R i (j )
20
30
90
40
PM *
180
200
270
2
10
specs
10
!c*
10
frequency (rad/s)
50
60
3
10
10
P
R a P
L = R i R a P
2
10
10
frequency (rad/s)
P (jc ) + < P M
we need to obtain a phase lead of at least
P M P (jc ) +
R i (j )
Phase (deg)
magnitude: attenuation of |P (jc )|dB

phase: since
R a (j)
30
actions needed:
10
PM PM *
50
10
90
180
200
P
R a P
L = R i R a P
270
3
10
10
10
frequency (rad/s)
10
10
23
example we need an attenuation of 8 dB and a phase lead of at least 25

we need to use both lead and a lag compensators but in the proper order
we choose the lead compensator first in such a way to obtain a phase increase of the
required 25 plus an extra (for example of 8) in order to compensate the lag that will be
introduced by the lag compensator
This lead function will also introduce, at the chosen frequency, an amplification of exactly
a (j )|dB
|R
c
the lag compensator will be chosen so to introduce an attenuation of

a (j )|dB
8 dB + |R
c
and a lag smaller than the extra 8 previously introduced

ma = 8
! = 0.8
34
2.5 dB
mi = 3.2
! = 20
-10.5 dB
< 8
(these numbers are just for illustration purposes and have not been verified)

24
PID controllers
3 basic heuristic actions
Proportional: the control action is set to be directly proportional to the system error (present)
Integral: the control action is set to be proportional to the system error integral (past)
Derivative: the control action is set to be proportional to the system error derivative (future)
output of the
controller
e (s)
m(t) = KP e(t) + KI
m (s)
CPID (s)
e( )d + KD e(t)
m(s)
CP ID (s) =
e(s)
ideal PID controller

transfer function
KI
KP +
+ KD s
s
1
KP 1 +
+ TD s
TI s
with
KP
TI =
KI
TD
KD
=
KP
widely spread
fixed structure with 3 tunable parameters KP KD KI
can be tuned automatically even with scarce knowledge of the (simple) plant
basic tuning refers only to static specifications

25
Derivative action is physical not realizable (improper transfer function) we need to approximate
add a high frequency pole in
KP
N
s=
N =
KD
TD
a
CD
(s)
TD s
=
TD
1+ N s
high frequency pole
Magnitude (dB)
60
40
tive
a
v
i
r
l de
a
e
d
i
20
0
approx
20
N = 10, 20, 50, 100
40
2
10
10
10
10
frequency (rad/s)
10
10
Phase (deg)
ideal derivative
approximated
derivative action
for various values of N
90
45
N = 10, 20, 50, 100
0
2
10

10
10
10
frequency (rad/s)
10
10
26
CaPID (s)
1
r
1
KI s
+
+ +
m
KP
P (s)
TD s
TD
1+ N s
Basic PID feedback control scheme

typical configurations
proportional
proportional + derivative (approximate)
proportional + integrative

27
PD configuration
30
CPa D (s)
= KP
sTD
1+
1 + s TND
= KP
1 + TD NN+1 s
N +1
1
+ N +1 TD N
1/ma
Magnitude (dB)
20
10
0
10 2
10
Phase (deg)
equivalent to a Lead compensator
KP = TD = 1, N = 10
1
1+
sTI
KP (1 + TI s)
=
TI
s
Magnitude (dB)
CP I (s) = KP
10
10
10
10
frequency (rad/s)
10
10
10
KP = TI = 1
20
0
2
10
10
10
10
10
frequency (rad/s)
10
10
10
frequency (rad/s)
10
10
10
Phase (deg)
10
10
10
frequency (rad/s)
40
zero in s = -1/TI + pole in s = 0

compensates the phase lag introduced by the pole in s = 0
45
0 2
10
PI configuration
10
28
Vocabulary
English
Italiano
lead compensator
funzione anticipatrice
lag compensator
funzione attenuatrice
phase lead
anticipo di fase
phase lag
ritardo di fase
attenuation
attenuazione
amplification
amplificazione
open-loop shaping
sintesi per tentativi

29

2014 Lec 16 Loop Shaping

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

2014 Lec 16 Loop Shaping

Загружено:

Авторское право:

Доступные форматы

Control Systems

Control Design: Loop shaping

Sunday, November 30, 2014

Lanari: CS - Loop shaping

location of eigenvalues/poles in the complex plane

location of eigenvalues/poles in the complex plane

Lanari: CS - Loop shaping

resonant peak Mr (and overshoot Mp)

Lanari: CS - Loop shaping

necessary part of the controller

to be still chosen in order to satisfy the dynamic specifications

Lanari: CS - Loop shaping

Open Loop Shaping

with some desired

C (s) = C1 (s)R (s)

dynamic performance specs

Lanari: CS - Loop shaping

P (s) = C1 (s)P (s) = h P (s)

Amplification: we need to increase the magnitude at some frequency

Lead: we need to increase the phase at some frequency

so if we have extra phase we can keep it

2 possible actions on the

maximum lag we can tolerate

2 possible actions on the

Lanari: CS - Loop shaping

zero in s = 0 to obtain a phase lead

both have unit gain so no magnitude change in ! = 0

Lanari: CS - Loop shaping

Lanari: CS - Loop shaping

amplification or attenuation at !c*

choose ma (or mi) and the normalized frequency !

Lanari: CS - Loop shaping

for different values of ma

for Ri(s) just change sign

Lanari: CS - Loop shaping

frequency which coincides with the desired one

Lanari: CS - Loop shaping

magnitude: as small amplification as possible in

Sunday, November 30, 2014

example we need a phase lead of 25 with the smallest amplification possible

Lanari: CS - Loop shaping

0.5 is the smallest (for

with the amplification of

several choices of ma are

we can tolerate at most a lag of

Lanari: CS - Loop shaping

example we need an attenuation of 17 dB and can tolerate a maximum lag of 7

Sunday, November 30, 2014

60 is the smallest (for

Lanari: CS - Loop shaping

maximum allowed lag

together with a lag

several choices of mi are

on the choice of the normalized frequency (case II)

starts attenuation before strictly needed

Lanari: CS - Loop shaping

both compensators solve the specifications

we need to obtain a phase lead of

Lanari: CS - Loop shaping

example we need an amplification of 20 dB and a phase lead of at least 25

usually requires a choice of the normalized frequency on the right-hand side

Lanari: CS - Loop shaping

example we need an amplification of 20 dB and a phase lead of at least 25

with the amplification of