Lecture 9 - Specification and Assessment Introduction

K. J. Åström • Important to specify and evaluate systems

• Follow reference signals
1. Introduction
• Reduce load disturbances
2. Classical Specifications
• Do not inject too much measurement noise
3. Properties of Simple Transfer Functions
• Sensitivity to modeling errors
4. Limitations
• Limitations
5. Assessment
• Quantitative Expressions
6. Summary
– Time and frequency domains
7. Appendix - Examples Simple Transfer Functions
• Many classical specifications were geared towards re-
Theme: To specify and evaluate. Important particularly if you sponse to reference signals.
are a customer of control. • Important to consider response to disturbances.

Some Key Issues Classical Specifications

d n
• Focused on one a few of the transfer functions, typically
r u x y response to step in reference signal
F ∑ C ∑ P ∑
• Specifications in both time and frequency domains
• Some robustness measure, e.g. gain and phase margins
−1
• Remember that a few transfer functions only give a limited
view. You must look at all six transfer functions to really
• Six transfer functions should be considered know how the system behaves
• Transient responses (steps) • Still useful to know about this because of the tradition.
This is what many persons you will encounter will talk
• Nyquist diagrams
about.
• Bode diagrams
• Poles and zeros

Different representations for different purposes

&
c K. J. Åström August, 2000 1
 Response to Step Reference Signal Response to Ramp in Reference Signal
10 10
Useful parameters
• Overshoot M
• Rise time Tr 5 5

• Solution time Ts
• Final error e0 0 0
0 2 4 6 8 10 0 2 4 6 8 10
Necessary to look at control signal required to generate the
step! • Useful to describe tracking of slow signals.
• Particularly important in motion control

Response to Step in Load Disturbance Frequency Response to Reference Signal

Control error for unit step load disturbance

0.4

0.2

0 2 4 6 8 10

How to measure it?

• Max error
Useful parameters
• Max time
• Solution time • Bandwidth
• Integrated absolute error IE • Resonance peak
• Integrated error IE
&
c K. J. Åström August, 2000 2
 Nyquist Curve of Loop Transfer Function Bode Plot of Loop Transfer Function
Useful parameters 10
1 Gain
• Gain margin
• Gain crossover fre- 10
0

quency ω g c
• Phase margin ϕ m −1
10

• Phase crossover fre-

−1 0
10 10

quency ω 180 Phase

−100
• Max sensitivities Mt , Ms −120

• Frequencies ω t, ω s −140

−160

−180

−200
−1 0
10 10

Poles and Zeros of a Transfer Function More Complete Specifications

• Make specifications on all six transfer functions
• Several choices, one possibility:
• Show output and control signal in response to a step in
the reference signal F PC/(1 + PC), FC/(1 + PC)
• Show the response of process output to step change in
load disturbance P/(1 + PC)
• Show Bode diagram of transfer function from measure-
ment noise to control signal − C/(1 + PC), (Skip the minus-
sign!)
• Show Bode diagrams of the sensitivity function S =
1/(1 + PC) and the complementary sensitivity function
T = PC/(1 + PC).

&
c K. J. Åström August, 2000 3
 Properties of Simple Systems Second Order System with a Zero
• It is necessary for you to have a working knowledge
ω 20 s+a ω 20 1 sω 20
of properties of simple systems. We will give a short G ( s) = = 2 +
a s2 + 2ζ ω 0 s + ω 0
2
s + 2ζ ω 0 s + ω 0
2 a s + 2ζ ω 0 s + ω 20
2
discussion here.
• Practice this in the home works Notice that G (0) = 1. Let h0 be the step response of

• Use Matlab and interactive software ω 02

G0 =
• Many results are collected in an appendix to this lecture. s2 + 2ζ ω 0 s + ω 02
See slides at the end. . The step response of G is then
1 dh0 ( t)
h( t) = h0 ( t) +
a dt
• Implications for the shape of the response
• Effect of different values of a
• What happens if a is negative?

Second Order System with a LHP Zero Second Order System with a RHP Zero
Step responses for ω 0 = 1 and ζ = 0.5 a = 0.25, 0.5 1, 100 Step responses for ω 0 = 1 and ζ = 0.5 a = −0.25, -0.5 -1,
Step Response
-100
3 Step Response

1.5

2.5
1

2
0.5
Amplitude

1.5 Amplitude 0

−0.5
1

−1
0.5

−1.5

0
0 5 10 15
Time (sec.) −2
0 5 10 15
&
c K. J. Åström August, 2000 Time (sec.) 4
 Limitations Non-minimum Phase Systems
There are many factors that limits the achievable performance Factor process transfer function as P(s) = Pmp(s) Pnmp(s)
where such that e Pnmp(iω )e = 1 and the sign chosen to give
• Nonlinear effects such as saturation and rate saturation
negative phase of Pnmp, then
• Measurement noise
arg L( iω g c ) = arg Pnmp( iω g c ) + arg Pmp ( iω g c) + arg C( iω g c) ≥ −π + ϕ m
• Dynamics with non-minimup phase characteristics, time
delays, poles and zeros in the right half plane
Bode’s relations give arg Pmp(iω g c ) + arg C(iω g c )  nπ /2 where
• Philosophy: n is the slope of the gain curve of the loop transfer function at
– Understand the limitations the gain crossover frequency. The inequality then becomes
– If at all possible modify the process π
arg Pnmp( iω g c ) ≥ −π + ϕ m − n
– In all cases do not pose unrealistic specifications 2
Simple Rule of Thumb ϕ m = 45 and ng c = −0.5 gives
○

π
arg Pnmp ( iω g c) ≥ −
2

System with RHP Zero System with Time Delay

Assume
Pnmp ( s) = e− sT
z− s
Pnmp ( s) =
z+ s Hence
Hence
ω arg Pnmp( iω g c ) = −ω g c T
arg Pnmp(iω g c ) = −2 arctan
z Cross over frequency inequality
Cross over frequency inequality
π
ω gc π ω g c T ≤ π − ϕ m + ng c
arg Pnmp(iω g c) = −2 arctan ≥ −π + ϕ m − ng c 2
z 2
Choosing ϕ m = π /4 and ng c = −1/2 gives
Hence
π
ω gc
≤ tan(
π
−
ϕm π
+ ng c )
ω gcT ≤
z 2 2 4 2
Choosing ϕ m = π /4 and ng c = −1/2 gives ω g c < z

&
c K. J. Åström August, 2000 5
 System with RHP Pole System with RHP Pole and Zero Pair

s+ p ( z − s)( s + p)
Pnmp( s) = Pnmp( s) =
s− p ( z + s)( s − p)
Hence
p For z > p we have
arg Pnmp(iω g c ) = −2 arctan
ω arg Pnmp ( iω ) = −2 arctan
ω g c / z + p/ω g c
Cross over frequency inequality 1 − p/ z
p π Cross over frequency inequality
−2 arctan ≥ −π + ϕ m − ng c
ω gc 2 ω gc p p π ϕm π
+ ≤ (1 − ) tan − + ng c
Hence z ω gc z 2 2 4
p Simple calculations give
ω gc ≥
tan(π /2 − ϕ m /2 + ng cπ /4) √
z 2 + 2 1 +α2
≥1+
Choosing ϕ m = π /4 and ng c = −1/2 gives ω g c ≥ p p α2

RHP Pole-Zero Pair Example - The X-29

√ Advanced experimental aircraft. Much design effort was done
z 2 + 2 1 +α2
p
≥ 1+
α2 with many methods and much cost. Specifications ϕ m = 45○
could not be reached. Here is why!
Assume a phase margin ϕ m = π /4 and a slope at the
crossover frequency of ng c = −1/2, then Non-minimum phase part of the transfer function
s − 26
z ≥ 5.83p Pnmp( s) =
s−6
Phase margin The zero pole ratio is z/ p = 4.33 with ng c = −1/2 we get
r
π p ϕ m = 32.4
ϕ m < π + ng c − 4 arctan
2 z
A phase margin of 45○ cannot be achieved!
z/ p 2 2.24 3.86 5 5.83 8.68 10 20
ϕm -6.0 0 30 38.6 45 60 64.8 84.6

&
c K. J. Åström August, 2000 6
 Example - Klein’s Unrideable Bike Summary of Limitations - Part 1
Bicycle with rear wheel steering • A RHP zero z

θ ( s) mhV0 V0 − bs ω gc 0.5 for Ms, Mt < 2
= ≤
δ ( s) J p c s 2 − mg h / J p z 0.2 for Ms, Mt < 1.4.

Typical values: • A time delay T

RHP zero and pole at 
m = 70 kg 0.7 for Ms , Mt < 2
s=z=
V0 ω gcT ≤
h = 1.2 m 0.37 for Ms , Mt < 1.4.
xf
s x f = 0.7 m
mg h • A RHP pole p
s=p= J p = 120 kgm3
Jp 
V0 = 5 ms−1 p 2 for Ms , Mt < 2
Hence ≥
ω gc 5 for Ms , Mt < 1.4.
z mg hV0 ○
= = 2.74, ; ϕ m = −10.4
p Jpx f

Summary of Limitations - Part 2 Assessment of a Control System

• A RHP pole-zero pair with z > p • If you are designing control system it is important to
 assess their properties
z 6.5 for Ms, Mt < 2
≥ • If you are a customer to control system designers is is
p 14.4 for Ms, Mt < 1.4.
useful to assess their properties
• A RHP pole-zero pair with z < p • If you are using control systems it is important to assess
 their performance
p 6.5 for Ms , Mt < 2
≥ • You must consider all properties not just the response to
z 14.4 for Ms , Mt < 1.4
reference signals
• A RHP pole p and a time delay T

0.16 for Ms , Mt < 2
pT ≤
0.05 for Ms , Mt < 1.4.

&
c K. J. Åström August, 2000 7
 Examples PI Control of a First Order System
Previously we have seen that it was straight forward to use b
Processmodell: P =
pole placement to design a controller for a given process. We s+a
use our new insight into the properties of the basic feedback ki
loop and specifications to make an assessment of the design. Controller: U = − kY + ( R − Y )
s
Two simple examples are used as illustration.
b( ks + ki)
• PI Control of a first order system Loop transfer function: L =
s ( s + a)
• PI Control of a second order system Characeristic equation: s2 + s(a + bk) + bki = 0
Choose k = (2ζ ω 0 − 1)/b and ki = ω 02 /b to give s2 + 2ζ ω 0 s +
ω 02 = 0.
Are there any restrictions on ζ and ω 0 ?

The Gang of Six The Sensitivity Function

Response to reference signal Skectch the Bode plot for the sensitivity function
ω 20 ω 20 ( s + a) s( s + a)
Y ( s) = R( s), U ( s) = R( s S( s ) =
s2 + 2ζ ω 0 s + ω 20 b( s2 + 2ζ ω 0 s + ω 20 ) s2 + 2ζ ω 0 s + ω 20

Response to load disturbance for a = 1, ζ = 0.5 and ω 0 = 0.1

bs
Y ( s) = D ( s)
s2 + 2ζ ω 0 s + ω 20

Response to measurement noise

( s + a)( ks + ki ) ( s + a)((2ζ ω 0 − a) s + ω 20 )
U ( s) = − N ( s) = − N ( s)
s + 2ζ ω 0 s + ω 0
2 2 s2 + 2ζ ω 0 s + ω 20

Sensitivity functions The Audience is Thinking ...

s( s + a) ( ks + ki ) b ((2ζ ω 0 − a) s + ω 20 )
S( s ) = , T ( s) = =
s2 + 2ζ ω 0 s + ω 20 s2 + 2ζ ω 0 s + ω 0
2 s2 + 2ζ ω 0 s + ω 20
&c K. J. Åström August, 2000 8
 The Sensitivity Function The Complementary Sensitivity Function
Bode plot for ζ = 0.7, a = 1 and ω 0/a = 0.1, 1 and 10 Bode plot for ζ = 0.7, a = 1 and ω 0 /a = 0.1, 1 and 10
Bode Diagrams Bode Diagrams

20 20

0
0

−20
−20
−40
Phase (deg); Magnitude (dB)

Phase (deg); Magnitude (dB)

−40
−60

−80 −60

150 0

−50
100
−100

50 −150

−200
0
−250

−50 −300
−2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10

Frequency (rad/sec) Frequency (rad/sec)

Choosing to small ω 0 gives a system that has poor robustness. Choosing to small ω 0 gives a system that has poor robustness.

Conclusion Control of a Thermal Process

Process model: P(s) = s+b a Design a PI controller to give a You are in charge of a project concerning temperature control
closed loop system with the characteristic polynomial for a wafer production. A simple process model is
s2 + 2ζ ω 0 s + ω 20 1
P( s ) =
( s + 1)( s + 0.02)
2ζ ω 0 − a ω 20
Controller parameters: k = , ki =
b b An engineer that you supervised has proposed
To obtain reasonable robustness the parameter ω 0 cannot be 1
too large, (ω 0 = xxxa gives sensitivities less than 2. U = − Y (+ (R − Y)
50s
Too high values gives a system that is too sensitive to mea- He said that the response to a step in the reference signal
surement noise k = 2ζ ω 0 ! originally was quite high but that he eliminated it by set point
weighting. He shows you some very nice step responses
Even if we can place the poles arbitrarily the process imposes
to changes in the reference signal. He also shows that the
limitations!
maxmal values of the sensitivity functions are reasonable. Has
he given you a good controller?

&
c K. J. Åström August, 2000 9
 The Gang of Six Response to Step in Load Disturbance
Response to reference signal 2
Process output
1 ( s + 1)( s + 0.02) 1.5

Y ( s) = R( s), U ( s) = R( s
s2 + s + 1 s2 + s + 1 1

0.5
Response to load disturbance
0

s + 0.02
0 20 40 60 80 100 120 140 160 180 200
s
Y ( s) = D ( s) U ( s) = − 2 D ( s) Control signal
( s + 0.02)( s2 + s + 1) s +s+1 0

−0.2

Response to measurement noise −0.4

−0.6

( s + 0.02)( s + 1)
−0.8

U ( s) = − N ( s) −1

s2 + s + 1 −1.2

−1.4
0 20 40 60 80 100 120 140 160 180 200

Sensitivity functions
s( s + 1 ) 1 Very slow recover of load disturbance! Why does the control
S( s ) = , T ( s) =
s2 + s + 1 s2 +s+1 signal not eliminate it?

Drastically Improved Control Drastically Improved Control

Changing Ti from 50 to 4 gives a drastic improvement • Specifications are important
2
Process output • Classical specifications
1.5 • Important to consider the “Gang of Six”
1
• Improved specifications
0.5

0
• Limitations. Non-minimum phase properties are bad! Try to
−0.5
redesign the process!
0 20 40 60 80 100 120 140 160 180 200

• Assessment. Things to think about when to are in the

0
Control signal market for control!
−0.2

−0.4

−0.6

−0.8

−1

−1.2

−1.4
0 20 40 60 80 100 120 140 160 180 200

&
c K. J. Åström August, 2000 10
 Summary Appendix - Properties of Simple Systems
• Classical specifications do not cover all In this Appendix we have collected properties of simple sys-
• Important to look at all six transfer functions tems
• Properties of simple transfer functions • First order systems
• Non-minimum phase elements give severe limitations • Second order systems without zero
– Time delays • Third order system without zeros
– RHP zeros • Second order system with zero
– RHP poles
• Examples of assessment of a control system

Second Order System without Zeros Step Response

Transfer function 

 e−ζ ω 0 t
ω 20 
 1 − p sin (ω d t + φ ), ζ <1

 1 −ζ2
G ( s) = 

s2 + 2ζ ω 0 s + ω 20
h( t) = 1 − (1 + ω 0 t) e−ζ ω 0 t, ζ = 1


Introduce 
  

 ζ
p 
1 − cosh ω d t + p 2 sinh ω d t e−ζ ω d t
ωd = ω0 1 −ζ2 ζ −1

Poles where
 p
−ζ ω 0 ± iω 0 1 − ζ 2 = −ζ ω 0 ± iω d if ζ ≤ 1 φ = arccos ζ
s=
−ζ ω ± ω pζ 2 − 1 = −ζ ω ± ω if ζ > 1
0 0 0 d

p p
where ω d = ω 0 1 − ζ 2 if ζ ≤ 1 and ω d = ω 0 ζ 2 − 1 if ζ > 1

&
c K. J. Åström August, 2000 11
 Step Response Frequency Response
ζ = 0, 0.1, 0.2, 0.5, 0.707, 1, 2, 5 and 10. Bode plot for ζ = 0.1, 0.2, 0.5, 0.707, 1, 2, 5 och 10.
Step Response Bode Diagrams

2
20
1.8
0
1.6
−20
1.4

Phase (deg); Magnitude (dB)

−40
1.2
Amplitude

−60
1
0
0.8
−50
0.6
−100
0.4

−150
0.2

−200
−1 0 1
0 10 10 10
0 5 10 15
Time (sec.) Frequency (rad/sec)

Third Order System without Zeros Third Order System without Zeros
Transfer function Stgep responses for ζ = 0.5 and a/ω 0 =0.25, 0.5, 1, 2 and 10.
aω 20 Step Response
G ( s) =
( s + a)( s2 + 2ζ ω 0 s + ω 20 )

Notice G (0) = 1. 1

What do we mean by a solution?

0.8

step([a],conv([1 a],[1 2*z*w0 w0^2]))

Amplitude 0.6
ICTools and SysQuake!
0.4

0.2

0
0 5 10 15
Time (sec.)

&
c K. J. Åström August, 2000 12
 Third Order System without Zeros Second Order System with a Zero
Bode plots for ζ = 0.5 and a/ω 0 =0.1, 0.2, 0.5, 1, 2, 5, 10
ω 20 s+a ω 20 1 sω 20
G ( s) = = 2 +
Bode Diagrams
a s2 + 2ζ ω 0 s + ω 0
2
s + 2ζ ω 0 s + ω 0
2 a s + 2ζ ω 0 s + ω 20
2

Notice that G (0) = 1. Let h0 be the step response of

50

ω 02
−50
G0 =
s2 + 2ζ ω 0 s + ω 02
Phase (deg); Magnitude (dB)

−100

−150 . The step response of G is then

0
1 dh0 ( t)
−50 h( t) = h0 ( t) +
−100
a dt
−150
• Implications for the shape of the response
−200

−250 • Effect of different values of a

• What happens if a is negative?
−300
−2 −1 0 1 2
10 10 10 10 10

Frequency (rad/sec)

Second Order System with a LHP Zero Second Order System with a RHP Zero
Step responses for ω 0 = 1 and ζ = 0.5 a = 0.25, 0.5 1, 100 Step responses for ω 0 = 1 and ζ = 0.5 a = −0.25, -0.5 -1,
Step Response
-100
3 Step Response

1.5

2.5
1

2
0.5
Amplitude

1.5 Amplitude 0

−0.5
1

−1
0.5

−1.5

0
0 5 10 15
Time (sec.) −2
0 5 10 15
&
c K. J. Åström August, 2000 Time (sec.) 13
 Second Order System with a LHP Zero Second Order System with a RHP Zero
Bode plot for ζ = 0.5 and a/ω 0 =0.1, 0.2 0.5 1 2 5 10 Bode plot for ζ = 0.5 and a/ω 0 =-0.1, -0.2 -0.5 -1 -2 -5 -10
Bode Diagrams Bode Diagrams

20 20

0 0

−20 −20
Phase (deg); Magnitude (dB)

Phase (deg); Magnitude (dB)

−40 −40

−60 −60

100 0

50 −50

−100
0
−150
−50
−200
−100 −250

−150 −300
−2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10

Frequency (rad/sec) Frequency (rad/sec)

&
c K. J. Åström August, 2000 14