Analysis and Design of Linear Control Systems, 04th.

3rd July, 4th Lecture

Analysis and Design of Linear Control System Part2Spring, 2012 Instructor: Prof. Masayuki Fujita (S5-305)
Lecture Time and Place: Tuesday, 10:4512:15, S224 Schedule: 3rd, 10th, 24th, 31st, July TA: Tatsuya Ibuki, Yasuaki Wasa(S5-302)

11 Frequency Domain Design

11.5 Fundamental Limitations Keyword : Right Half-Plane Poles and Zeros Gain Crossover Frequency Inequality

Recap. Gain Crossover Frequency Inequality

Factor the process transfer function as

P ( s ) = Pmp ( s ) Pap ( s )
L( j )
ngc : slope at gc

Pmp : minimum phase part

Pap : all-pass system

[Ex. 11.7] Zero in the right half-plane All-pass system with a RHP zero

Im

Pole Zero

Pap ( s ) =

gc : gain crossover freq.

arg L( j )

arg Pmp

arg Pap ( j ) = {arg( z j ) arg( z + j )} =

Im
arg( z + j ) = arctan

Phase lag of the all-pass system

zs z+s

z>0

Re

*RHP : right half-plane

arg P

arg Pap

m : required phase margin

arg( z j ) = arctan

Re

Gain Crossover Frequency Inequality

arg Pap ( j gc ) m + n gc

=: l

(11.15)

gain crossover frequency inequality arg Pap ( j gc ) m + n gc =: l 2 Bound on the crossover frequency gc

(11.15)

allowable phase lag of Pap at gc : l

gc < z tan( l / 2 )

(11.16)

[Ex. 11.7] Zero in the right half-plane Bound on the crossover frequency gc
gc < z tan( l / 2 ) (11.16)
L ( j ) z tan( l / 2 )

Time Delay Exercise

gc < 0 . 6 z
better

l = 60
Im
0
worse

u (t )

U ( s)

e s

Y ( s)

u (t )

gc

(0.6z)

limitation
arg L( j )

Z
Step Response

Re

0
Gain [dB]

t
Bode Diagram
10
0

y (t ) = u (t ) t

Nyquist Plot Im

ej

Time [s]

Slow RHP zeros (z small) Fast RHP zeros ( z large)

Tight restrictions Loose restrictions

10 0 20 40 60 2 10

Phase []

1 Re

[rad/s]

10 1

100

Analysis and Design of Linear Control Systems, 04th.

Pade approximation Time delay Pade approximation

Im
0

worse

better

1 0. 5 s 2 / s = 1 + 0 .5 s 2 / + s

Re

[Ex. 11.8] Pole in the right half-plane All-pass system with a RHP pole

arg( j p ) = 90 + arctan

s+ p Pap ( s ) = s p

arg( p + j ) = 90 arctan

p>0
p

Im

Phase lag of the all-pass system

Re

Time delays also impose limitations similar to those given by zeros in the RHP.
A long time delay is equivalent to a slow RHP zero z = 2 /
L ( j )

arg Pap ( j ) = {arg( p + j ) arg( j p )} =

gain crossover frequency inequality arg Pap ( j gc ) m + n gc =: l 2

gc < z tan( l / 2 ) =

tan( l / 2 )

gc

z tan( l / 2 )

(1.2 / )

(11.15)

Exercise

l = 60

gc <

1 .2

arg L( j )

Bound on the crossover frequency gc p gc > (11.17) tan( l / 2 )

[Ex. 11.8] Pole in the right half-plane

Loop Shaping

Bound on the crossover frequency gc

gc >
p tan( l / 2 )
(11.17) L ( j )

Exercise

l = 60

gc > 1 . 7 p

L ( j )

Performance
L ( j ) >> 1
1 .2

gc
p tan(l / 2)

0. 6 z

(1.7p)
arg L( j )

1.7 p

Im

gc
Robust Stability

better

worse

L ( j ) << 1

0
p

Re

RHP zero Fast RHP poles ( p large) Slow RHP poles ( p small) Tight restrictions Loose restrictions Time Delay RHP pole

gc < z tan( l / 2 )

gc < z tan( l / 2 ) =
gc >
p tan( l / 2 )

tan( l / 2 )

Loop Shaping
gc
gm
L( j )
A
MS < 2

S ( j )

-3 bS

Right Half-Plane Poles and Zeros and Time Delays For systems with a RHP pole p and RHP zero z (or a time delay ), any stabilizing controller gives sensitivity functions with the property
M S = sup S ( j )

pc

L( j )
M T < 1.25 (2 [dB])

p+z pz

M T = sup T ( j ) e p

(11.18)

Fig. 11.8

bT
T ( j )
MS < 2

3[dB]

RHP pole and zero and time delay significantly limit the achievable performance of a system
MS
S( j) MT T ( j)

Gain Margin Phase Margin

g m = 1 / L(i pc )
(2 5) m = + arg L(i gc )

(30 60)

Stability Margin sm = 1 / M s (0.5 0.8)

M T < 1.25 MS 1 1 gm m 2 arcsin M S 1 2M S M S 1 1 1 m 2 arcsin gm 1+ MT 2M T M T

Analysis and Design of Linear Control Systems, 04th.

Right Half-Plane Poles and Zeros and Time Delays

Right Half-Plane Poles and Zeros and Time Delays The product of RHP pole and time delay must be sufficiently small

If RHP pole and zero are equal ( p = z ), there will be an unstable subsystem that is neither reachable nor observable, and the system cannot be stabilized
The zeros and the pole must be sufficiently far apart Ex. )
all-pass system

Ex. )

all-pass system

Pap ( s ) =

allowable phase lag of Pap at gc l = 90

bs s 1

* p =1

Pap ( s ) =

z=b

allowable phase lag of Pap at l = 90

e s s 1

gc

arg Pap ( j gc ) l
(11.15) Fig. 11.13 (b) RHP pole and time delay

arg Pap ( j gc ) l
(11.15) Fig. 11.13 (a) RHP pole/zero pair

p < 0 .3

z / p > 6 or z / p < 1 / 6

[Ex. 11.9] Balance system (6.3)

[Ex. 11.9] Balance system (6.3)

gc >
2

Equations of motion

(M

(J + ml )&& ml cos &p& = & + mgl sin

2

&& = c p & ml sin & + F & ml cos + m )& p

(6.4)

p tan( l / 2 )

RHP pole p = 2 . 68 RHP zero z = 2 . 09

RHP zero can be eliminated The gain crossover frequency inequality (11.15) is based just on the RHP pole
Fig. 6.2 (b)
t

Transfer functions
from F to from F to p
H F = H pF = ml ( M t J t m l ) s 2 + mglM
2 2

l = 45 l = 60

gc > 6 . 47 ( gc > 2 . 4 p )

gc
p tan(l / 2)

gc > 4 . 56 ( gc > 1 . 7 p )

J t s 2 + mgl 2 s ( ( M t J t m 2 l 2 ) s 2 + mglM t )

*Mt = M +m

J t = J + ml 2

Im

H pF :

RHP pole p = 2 . 68 RHP zero z = 2 . 09

Re

Pole Zero

If the actuators have sufficiently high bandwidth, e.g. a factor of 10 above gc or roughly 10 Hz, then we can provide robust tracking up to this frequency p+z M S = sup S ( j ) sup S ( j ) 8 pz Ideally S ( j ) < 2

difficult to control robustly

[Ex. 11.11] X-29 aircraft available bandwidth sensors : 120 rad/s control processors : 30-40 rad/s actuators : 70 rad/s aerodynamics : 100 rad/s airframe : 40 rad/s
Fig. 1.3 Components of computer-controlled system

[Ex. 11.11] X-29 aircraft X-29 longitudinal dynamics

available bandwidth of the actuators that stabilize the pitch : a = 40 [rad/s]

desired bandwidth of the pitch control loop : 1 = 3 [rad/s]

Assume that the sensitivity function S ( s ) is given

S ( j ) =

M s 1

( 1 )

S( j) = Ms

(1 a )

Real physical systems have a multitude of limitations on available bandwidth

=3 = 40

Fig 11.15 (b) Sensitivity analysis

Analysis and Design of Linear Control Systems, 04th.

[Ex. 11.11] X-29 aircraft Assume

[Ex. 11.11] X-29 aircraft

L(s) / 2

a
Fig 11.5 (b)

Bodes integral

X-29 aircraft maximum achievable phase margin : 35 Boundaries for standard flight control specifications phase margin : 45 X-29 is difficult to control. X-29 longitudinal dynamics z 26 poles : p = 6 = 4 .3 p 6 zeros : z = 26 Im -6

log S ( j ) d =

=
0

log S ( j ) d

M s log d + (a 1 ) log M s 0 1 = p
1

Exercise

Ms = e

( p +1 ) / a

= 1.75 ( < 2 )

1 = 3 [rad/s] a = 40 [rad/s]

p = 6

Pole
Zero

26 Re

M S S(jgc) =
maximum achievable phase margin : 35

1 2 sin(m / 2)

z / p > 6 or z / p < 1 / 6 Desirable Condition

It is difficult to achieve the specifications, no matter how the controller is designed.

Bodes Integral Formula (11.5)

log S > 0 , S > 1 log S < 0 S <1

log S ( j ) d = pk

3rd July, 4th Lecture

11 Frequency Domain Design
11.5 Fundamental Limitations Keyword : Right Half-Plane Poles and Zeros Gain Crossover Frequency Inequality

(11.19)

pk : right half-plane poles

Waterbed Effect
0

S
log T > 0 , T > 1

log T ( j )

d =

1 zi

(11.20)

zi : right half-plane zeros

Better RHP zeros Fast (big) RHP poles Slow (small) Worse Slow (small) Fast (big)

log T < 0 T <1

S +T =1

slow

fast

10th July, 5th Lecture

10 PID Control
Keyword : PID Control, Ziegler Nichols Tuning

11 Frequency Domain Design

11.4 Feedback Design via Loop Shaping

Keyword : Lead and Lag Compensation