=
For the open loop case: ( )
ol ol ol ol ol
T T D G G D G D G + = + = +
/ / ( )
1
/ /
ol
T
ol ol ol ol
G
T T D G D G
S
G G G G
= = =
For the feedback controller case:
( )
1 ( )
cl
cl cl
cl
G G D
T T
G G D
+
+ =
+ +
After some manipulations: it is much smaller than that of the
open loop case.
1
1
cl
T
G
cl
S
GD
=
+
Lecture 4: Part 1: A First Analysis of Feedback
Filtered cases
The open loop transfer function is:
ol ol
T GD F =
Similar to the transfer function achieved in the case w/o filter.
Therefore the discussion on stability, sensitivity, etc. are almost the same.
Lecture 4: Part 1: A First Analysis of Feedback
The feedback transfer function is:
The filtered closed loop structure can realize the best properties of both the open-loop
and the unity feedback closed-loop cases.
The controller D_cl can be designed to regulate the system for the disturbance W and
The sensor noise V, while the filter F is designed to improve the tracking accuracy.
Filtered cases
1 1 1
cl cl
ol
cl cl cl
GD F HGD G
Y R W V
GD H GD H GD H
= +
+ + +
Lecture 4: Part 1: A First Analysis of Feedback
Sensitivity:
As the gain increases, approaches 1
Filtered cases
1.0
cl
T
F
S =
1
1
cl
T
G
cl
S
GD H
=
+
1
cl
T
cl
H
cl
GD H
S
GD H
=
+
cl
T
H
S
Need to have a good sensor
(low in noise, and very stable)
Lecture 4: Part 1: A First Analysis of Feedback
1. Basic Equations of Control
2. Control of Steady-State Error to Polynomial Inputs: System Type
Steady-state error
0 0
( )
lim ( ) lim ( ) lim
1 ( )
ss
t s s
cl
sR s
e e t sE s
G s D
= = =
+
The steady state error will depend on the input and loop transfer function
1
1
( )
( )
( )
m
i
i
cl n l
j
j
s z
k
G s D
s
s p
=
=
+
=
+
1
1
(1 )
( )
(1 )
i
j
m
z
i
cl n l
p
j
T s
K
G s D
s
T s
=
=
+
=
+
1/
i
z i
T z =
1/
j
p j
T p =
1
1
m
i
i
n
j
j
z
K k
p
=
=
=
=
+
=
+
=
+
2
0
0
/
lim
1 ( )
/
1 lim ( )
ss
s
cl
cl
s
sA s
e
G s D
A s
G s D
=
+
=
+
=
3
0
/
lim
1 ( )
ss
s
cl
sA s
e
G s D
=
+
=
order 1
0
1 lim ( )
0
1
ss
cl
s
A
e
G s D
A
=
+
= =
+
0
0
/
lim
1 ( )
/
lim /
1 /
ss
s
cl
v
s
v
A s
e
G s D
A s
A K
K s
=
+
= =
+
2
0
2
0
/
lim
1 ( )
/
lim
1 /
ss
s
cl
s
a
A s
e
G s D
A s
K s
=
+
= =
+
Lecture 4: Part 1: A First Analysis of Feedback
Steady-state error
t
( ) r t
A
0
Typical input signals
t
( ) r t
0
t
( ) r t
0
Step input Ramp input Parabolic input
A
s
2
A
s r At =
3
A
s
2
2
A
r t =
order 0
1
ss
p
A
e
K
=
+
ss
e =
ss
e =
order 1
0
ss
e =
/
ss v
e A K =
ss
e =
order 2
0
1 lim ( )
0
1
ss
cl
s
A
e
G s D
A
=
+
= =
+
0
/
lim
1 ( )
0
ss
s
cl
A s
e
G s D
=
+
=
2
0
/
lim
1 ( )
/
ss
s
cl
a
A s
e
G s D
A K
=
+
=
Lecture 4: Part 1: A First Analysis of Feedback
Steady-state error
Example 1: determine the system type and the relevant error constant for the speed control
e xample with proportional plus integral control having controller given by
cl p
D k = ( )
1
A
G s
s
=
+
( )
( )
1 1
p
cl p
Ak
A
G s D k
s s
= =
+ +
Lecture 4: Part 1: A First Analysis of Feedback
Type 0
Constant is:
p
Ak
Steady-state error
Example 2: determine the system type and the relevant error constant for the speed control
example with proportional plus integral control having controller given by
/
cl p i
D k k s = + ( )
1
A
G s
s
=
+
( )
( ) ( )
( ) (1 / )
( ) /
1 1 1
p i i p i
cl p i
A k s k Ak k s k
A
G s D k k s
s s s s s
+ +
= + = =
+ + +
Lecture 4: Part 1: A First Analysis of Feedback
Type 1
Constant is:
i
Ak
Steady-state error
Example 3: find the steady state error for the following system for unit step, ramp, and parabolic inputs
( 2)
( )
( 1)( 4)( 5)
cl
k s
G s D
s s s s
+
=
+ + +
2 (1 0.5 )
( )
20 (1 )(1 0.25 )(1 0.2 )
cl
k s
G s D
s s s s
+
=
+ + +
(1 0.5 )
( ) / 10
(1 )(1 0.25 )(1 0.2 )
cl
s
G s D k
s s s s
+
=
+ + +
Step input Ramp input Parabolic input
order 1
0
ss
e =
1/
1/( /10) 10/
ss v
e K
k k
=
= =
ss
e =
Lecture 4: Part 1: A First Analysis of Feedback
Steady-state error
Example 4: consider an electric motor position control problem including a non-unit feedback system caused
by having a tachometer fixed to the motor shaft and its voltage (which is proportional to shaft speed) is fed back
As part of the control. The parameters are:
1
( )
( 1)
( )
( ) 1
( ) 1
p
t
G s
s s
D s k
H s k s
F s
=
+
=
= +
=
=
1
( ) 1/
k
R s s
+
=
0
0 0
(1 ( ))
1
lim
, 1
t p
ss
k
s
p
k
T s
k k
e
k
s
k
+
= =
=
= =
In a similar way as for the reference inputs, the system is type 0 if a step disturbance input results in
a non-zero constant steady state error
The system is type 1 if a ramp disturbance input results in a steady-state value of the error that is a
non-zero constant, etc.
,
( ) ( )
n
w o w
T s s T s =
,
1
0 0
1
lim[ ( ) ] lim[ ( ) ]
n
ss w o w
k k
s s
s
y sT s T s
s s
+
> >
= =
If n>k, then he error is zero.
If n<k, the error is unbounded
If n=k, the system is type k
System type for regulation and disturbance rejection
Lecture 4: Part 1: A First Analysis of Feedback
Example 1: considering the simplified model of a DC motor in unity feedback shown below, where the
Disturbance torque is labeled W(s).
(a) Use the controller D(s)= kp
(b) Let the controller transfer function be given by:
(c) D(s) = kp+ki/s
The transfer function from W to Y is:
0
,
( ) ( 1)
( ) ( 1)
1
( 1)
o w
p
p
A
Y s B B s s
s T
A
W s A s s Ak
k
s s
+
= = =
+ +
+
+
n=0
,
0 0
( )
lim lim
( ) ( 1)
ss o w
s s
p p
Y s B B
e K
W s s s Ak Ak
= = = =
+ +
System type for regulation and disturbance rejection
Lecture 4: Part 1: A First Analysis of Feedback
(b) The transfer function from W to Y is:
( )
2
( ) ( 1)
( ) ( 1) ( )
1 /
( 1)
p I
p I
A
Y s B Bs s s
A
W s A s s A k s k
k k s
s s
+
= =
+ + +
+ +
+
n=1
2
0
1
lim
( 1) ( )
ss
s
p I i
Bs B
e
s s A k s k s Ak
= =
+ + +
Step input:
Truxals formula (1955) for the Error constants
Lecture 4: Part 1: A First Analysis of Feedback
Describes a relationship between the velocity constants of a Type 1 system in terms of
the closed-loop poles and zeros, a formula that connects the steady-state error to the systems
dynamic response.
1 2
1 2
( )( )...( )
( )
( )( )...( )
m
n
s z s z s z
T s K
s p s p s p
=
For type 1, the response to a step input is zero, therefore DC gain is 1:
( )
( ) ( ) ( ) ( )[1 ] ( )[1 ( )]
( )
Y s
E s R s Y s R s R s T s
R s
= = =
For type 1, system error due a unit ramp input is given by
2
1 ( )
( )
T s
E s
s
=
0 0
1 ( ) ( ) 1
lim lim
ss
s s
v
T s dT s
e
s ds K
= = =
This means: 1/Kv is related to the slope of the transfer function at origin.
(0) 1 T =
Truxals formula (1955) for the Error constants
Lecture 4: Part 1: A First Analysis of Feedback
Describes a relationship between the velocity constants of a Type 1 system in terms of
the closed-loop poles and zeros, a formula that connects the steady-state error to the systems
dynamic response.
0 0
1 ( ) ( ) 1
lim lim
ss
s s
v
T s dT s
e
s ds K
= = =
(0) 1 T = Because
| |
0 0
ln ( )
( ) 1
lim lim
ss
s s
d T s
dT s
e
ds T ds
= =
Substitute
1 2
1 2
( )( )...( )
( )
( )( )...( )
m
n
s z s z s z
T s K
s p s p s p
=
1 2
0
1 2
0
1 1
( )( )...( )
lim ln
( )( )...( )
lim ln( ) ln( )
m
ss
s
n
m m
i i
s
i i
s z s z s z d
e K
ds s p s p s p
d
K s z s p
ds
= =
=
`
)
= +
`
)
Truxals formula (1955) for the Error constants
Lecture 4: Part 1: A First Analysis of Feedback
1 2
0
1 2
0
1 1
( )( )...( )
lim ln
( )( )...( )
lim ln( ) ln( )
m
ss
s
n
m m
i i
s
i i
s z s z s z d
e K
ds s p s p s p
d
K s z s p
ds
= =
=
`
)
= +
`
)
0 0
1 ( ) ( ) 1
lim lim
ss
s s
v
T s dT s
e
s ds K
= = =
Since:
1 1
1 1 1
n m
i i
v i i
K p z
= =
= +
Truxals formula - example
Lecture 4: Part 1: A First Analysis of Feedback
A third-order type 1 system has closed-loop poles at -2+/-2j and -0.1. The system
Has one closed-loop zero. What should the zero be if Kv=10 is desired?
1 1 1 1 1
0.1
2 2 2 2 0.1
v
K j j z
= = +
+
0.0962 z =
Blank
The Three-Term Controller: PID controller
Lecture 4: Part 2: A First Analysis of Feedback
PID: ( )
I
p D
k
D s k k s
s
= + +
What Is the PID Controller Family?
P --- proportional
I --- integral
D --- derivative
PID is simple, it has weakness; it limits the range of plant that they can control successfully
--- not optimal
--- difficult in high order system
--- may be difficult in nonlinear systems control design, because it is not easier or impossible to
find poles/zeros
Control with an up-to-second-order controller
d
k dt
dt
Lecture 4: Part 2: A First Analysis of Feedback
Proportional gain
Provide a contribution depend on the instaneous value of the control error
Proportional gain can be used to control any stable plant
But it provides limited performance, e.g., larger transient stage
Non-zero steady state error, due to the fact that its frequency response is bounded
for all the frequencies.
p
k
e
p
k
( )
( )
= =
p p
s
k e k
E s
( ) M
So for very low and very high frequency, the transfer function gain is the same
Very small error => small => not enough to control the system
to have zero steady-state error.
X s
xdt
s
Lets look at this one:
I
k
r
1
s
P
e
x
( ) ( ) =
I
k
s E s
s
=
I
x k e
( ) ( ) =
I
k
X s E s
s
= x
[0] [ ] = +
I
x x k e
[1] [0] = + x e
Lecture 4: Part 2: A First Analysis of Feedback
Integral gain
I
k
s
How does this control affect the response
If the response has steady-state error, then integral of error increases, so is increasing to eliminate error.
If the response has no steady-state error, then integral in constant, so is does not change.
u
y
t
edt
t t
u
y
t
edt
t t
dB
Problem 1: its pole at origin will introduce stability problem
Problem 2: Integral of errors -> control action -> actuator saturation -> wind up
Problem 3: cannot affect all model response characteristic even combined with proportional gain
Lecture 4: Part 2: A First Analysis of Feedback
Derivative gain
=
D
sk e
Surface moves in proportional to the rate of change in error. How does this control affect the response?
If x increase fast, then rising time is small and overshoot will probably be big
If x increase slow, then rising time is large and overshoot will probably be small
Basically derivative control seeks to control rate of change of error. So it helps to balance the rising time
and overshoot by controlling speed of response
Limitation: It is tendency to yield large control signals in response to high frequency control errors, such
as errors induced by set-point changes or measurement noise. (why?)
So its implementation requires properness of the transfer functions. (n>=m) Also because of ?
So a pole is sometime added to the derivative
t t
x
x
1 +
D
D D
k s
k s
D
derivative time constant is normally chosen such that 0.1 0.2
D D D
k k
Lecture 4: Part 2: A First Analysis of Feedback
u
y
t
edt
t t
Derivative gain
=
D
sk e
See the high sensitive derivative term to noise
u
y
t
edt
t t
Noise
1 +
D
D D
k s
k s
So a pole is sometime added to the derivative, just like add a low pass filter
Noise Noise
=
D
sk e
1 +
D
D D
k s
k s
Lecture 4: Part 2: A First Analysis of Feedback
Figure 4.8 Block diagram of the PID controller: (a) with the D-term in the feedback path; and (b) with the D-term in the
forward path
Lecture 4: Part 2: A First Analysis of Feedback
Strength and weakness
P D I
Good
Bad
Simple, common sense
Like an amplifier
Need help to affect
Total response
Eliminate steady
States error
Limit stable range
So hurt stability (why?)
Good rate of response
Sensitive to noise
Usual effects of gain change
Gain Stability Speed Steady state error
increase
p
k
increase
I
k
increase
D
k
reduce
reduce
increase
Decrease
Eliminate
Small change
r s
t t
r s
t t
r s
t t
Lecture 4: Part 2: A First Analysis of Feedback
How to choose
p
k
I
k
D
k
Try and error
Use initial guess from theory (Ziegler Nichols), then fine tune using trial and error
Try and error
Strategy
check response using a gain;
change gain and check response
iterate
Advantages
Do not need to know theoretical model of system
Can use either computer or physical system
Disadvantages
Can be dangerous to a physical system
Can be time consuming
Can be hard to choose all gains
Some useful tips
Obtain an open-loop response and determine what needs to be improved
Add a proportional control to improve the rise time
Add a derivative control to improve the overshoot
Add an integral control to eliminate the steady-state error
Adjust each of gains until you obtain a final desired overall response
Keep the controller as simple as possible
Lecture 4: Part 2: A First Analysis of Feedback
P, PD, PID controller design example 1
Assume: we need to design a controller for
P
k
r T
T
e
3 2
1.151 0.1774
0.739 0.921
+
=
+ +
s
T
s s s
1
Control objectives
10%
2
4
2%
<
<
<
<
r
s
steady state
t s
t s
e
0.6
0.6
0.5
0.6 (0.125 )
=
=
=
p PU
PU
I
u
D PU u
k K
K
k
T
k K T
Attempts for basic response: magnitude of 2
nd
peak is size of 1
st
peak approximately
no guarantee of success but often gives reasonable point for control design that can be chang
trial and error iteration
Lecture 4: Part 2: A First Analysis of Feedback
Homework 5: textbook 4.2, 4.4, 4.8, 4.10, 4.17, 4.27, 4.32
Grading policy:
will check if you tried all of them or not (60%)
Will grade three problem if correct (40%)
-- if all wrong (24%)
-- if one correct (30%)
--- if two correct (36%)
Due on June 29
nd
in class (TA will collect the homework for me)
Lecture 4: A First Analysis of Feedback