linear feedback controller analysis - 21.

21. CONTROL SYSTEM ANALYSIS

Topics:

Objectives:

21.1 INTRODUCTION

21.2 CONTROL SYSTEMS

Control systems use some output state of a system and a desired state to make
control decisions.
In general we use negative feedback systems because,
- they typically become more stable
- they become less sensitive to variation in component values
- it makes systems more immune to noise
Consider the system below, and how it is enhanced by the addition of a
control system.

Control variable
INPUT
(e.g. gas)

vdesired

verror
+

OUTPUT
(e.g. velocity)

SYSTEM
(e.g. a car)

Driver or
cruise control

gas

car

vactual

The control system is in the box and could be a driver or a cruise control
(this type is known as a feedback control system)

Figure 21.1

An example of a feedback controller

Human rules to control car (also like expert system/fuzzy logic):

1. If verror is not zero, and has been positive/negative for a while, increase/decrease gas
2. If verror is very big/small increase/decrease gas
3. If verror is near zero, keep gas the same
4. If verror suddenly becomes bigger/smaller, then increase/decrease gas.
5. etc.

Figure 21.2

Rules for a feedback controller

 Some of the things we do naturally (like the rules above) can be done with
math- ematics

21.2.1 PID Control Systems

The basic equation for a PID controller is shown below. This function will try to
compensate for error in a controlled system (the difference between desired and actual
output values).

u
K c e Ki
edt

Figure 21.3

d
e

Kd
----dt

The PID control equation

The figure below shows a basic PID controller in block diagram form.

proportional

PID Controller

Kp e

V
e

+
-

Figure 21.4

Ki e

derivative
K d --d-- e
dt

+V

integral

amp

+
+

-V

A block diagram of a feedback controller

motor

e.g.
gas r
erro

c error

K
dt v

d dv error

K ---

Rules 2 & 3
(general difference)

Kc
Ki
Kd

---

dt
----------

Rule 4
(Immediate error)

Rule 1
(Long term error)

Relative weights of components

This is a PID Controller

Proportional
Integral
Derivative

For a PI Controller
K v
+ K i v error dt
gas = c error
For a P Controller
gas = K c v error
For a PD Controller
gas = K c v error

dv error
-------
+ K d ---
dt

The PID controller is the most common controller on the market.

21.2.2 Analysis of PID Controlled Systems With Laplace Transforms

1. We can rewrite the control equation as a ratio of output to input.

gas
error c error

= Kdt v

d
dv errordt

+K

-----
---------

d
gas
= K c + Ki dt + Kd -- --
-----------
dt
v erro r
Then do a Laplace transform
d x
--- -
sx dt

d
- -- s dt
1

dt

= s

a s
K
----L - g------ -- = K c + s i + Kd s
v error

xdt

x
= -s

The transfer function

2. We can also develop a transfer function for the car.

F = A gas = 10 ga s
F
---- ----- =

Transfer function for engine and transmission. (Laplace

transform would be the same as initial value.)

10
gas

= Ma =

F = M
--- d
v - ---

x
---dM 2
dt

d v
= M ---- dt

dt

v
L - -- =
F
1
--- ---Ms

Transfer function for acceleration of car mass

3. We want to draw the system model for the car.

gas

10

--1
--Ms

vactual

The system model is shown above.

If gas is specified directly, this is called open loop control. This is
not desirable, but much simpler.
The two blocks above can be replaced with a single one.

gas

1
- -0
--Ms

vactual

4. If we have an objective speed, and an actual speed, the difference is the

system error
v error = v desi red v act ual
set-point - desired system operating point

5. Finally, knowing the error is verror, and we can control gas (the control
variable), we can select a control system.
verror

Controller

K
----ga s = K c + i + Kd s
s
L ------------v erro r

gas

*The coefficients can be calculated using

classical techniques, but they are more
commonly approximated by trial and
error.

6. For all the components we can now draw a block diagram

vdesired

verror
+

Ki
K c + -----+ Kd s
s

gas

10

--1
---Ms

A negative feedback loop

that is the fundamental part of this
closed loop control system

vactual

21.2.3 Finding The System Response To An Input

output.

Even though the transfer function uses the Laplace s, it is still a ratio of input to

Find an input in terms of the Laplace s

vdesired
100

step
vdesired(t) = 100 for t >= 0 sec

ramp
vdesired(t) = 50t for t >= 0 sec
t(sec)

Input type

Time function

STEP

f t = Au t

RAMP

f t = Atu t

Laplace function
A
f s = --s
A
f s = - 2 --s
2

SINUSOID

f t = A sin t u t

PULSE

f t = A u t u t t1

etc......

= -----------f s s2 + 2
fs =

Therefore to continue the car example, lets assume the input below,
v desired t = 100

t 0 sec

1
v desired s = L v desi red t --0
-=
0
-s
Next, lets use the input, and transfer function to find the output of the system.
v
v act ual = act ual de sired
v
----------------v

desired

v actual

2
s K + s d K + K c100
i
--------------
= ------------------- -------------------------- -------s

2 M

- ---- + K + cs K i + K
s 10
d

To go further, some numbers will be selected for the values.

Kd = 10000
Kc = 10000
Ki = 1000
M = 1000

v actual

s 2 10000 + s 10000 + 1000

100
= ---------------------------------------------------------------------- -------

s 2 10100 + s 10000 + 1000 s

linear feedback controller analysis - 21.10

At this point we have the output function, but not in terms of time yet. To do this
we break up the function into partial fractions, and then find inverse Laplace
transforms for each.
2

2
s + s + 0.1
v act ual = 10 -------------------------------------------------
s s 2 1.01 + s + 0.1

Aside: We must find the roots of the equation, before we

can continue with the partial fraction expansion.
recall the quadratic formula,
2

ax + bx + c = 0

x =

b 4ac
x = -------------------------------2a
b
----

4 1.01 0.1 = 0.113 0.877

1 1
------- ------------------- -------------------

--- ----- --2 1.01 --

v ac t ual

2
10
s + s + 0.1

= ---------- --------------------------------------------------1.01 s s + 0.113 s + 0.877

A
B
C
v act u al = s --- s++---------+ ---------- ----------0.114 ----------s + 0.795

2 -- ----2 ---- ---- ----s

--------------------------------------------
=
-----
----------------
1
0
A = lim 0 .1 1 0 s0 .--1+---------------s +
s0

1.01

1.01 s s + 0.113 s + 0.877

0.113 0.877

A = 99.9

B =

2 --------------- ------- ----- --- ----- --- ------------------2 -- -----

s + 0.113

lim

s 0.113

1 0
s

1.01 s s + 0.113 s + 0.877

s +

0 .

1 20
2 + 0.113 + 0.1
-
= 0.264
0.113
B = - --- ---- 1.01 ------------------------------------------

--------------------- 0.113 0.113 + 0.877

C =

---------------- 2--- ------ 2

-------------------- s + 0.877
----- ---- ---- --- --

lim

s 0.877

1 1 0

1.01
s s + 0.113 s +
0.877

0 .

10 2 0.877 + 0.877 + 0.1

C

= ---------= 1.16

--------------------------------------------------------------- 0.877 0.877 + 0.113

1.01
0 .2 6 4
1 . 1 6
v act ual = -9--9--.--s9+- +
----- ---- --- ------ ------ --- -- --- ------0.113
s +- 0.877
s

Next we use a list of forward/inverse transforms to replace the terms in the

partial fraction expansion.
f s

f t

A
---2
s

At

Ae

A
--------- s+

Asin
t

n t

--s

A
----- --------- 2
2
s +
for 1
n 1
--------------------------------------2

sin n t 1

s + 2n s + n
etc.

To finish the problem, we simply convert each term of the partial fraction back to
the time domain.
99.9
0.264
1.16
v act u al = ---------- + --------------------- --------------------s
s + 0.113 s + 0.877

0.113t 1.16 e0.877 t

v actual = 99.9 + 0.264e

21.2.4 Controller Transfer Functions

The table below is for typical control system types,

Type

Transfer Function

Proportional (P)

Gc = K

Proportional-Integral (PI)

+
G c = K 1 -- 1
-s

Proportional-Derivative (PD)

G c = K 1 + s

Proportional-Integral-Derivative (PID)

G c = K 1 + --1-- + s

=
Lead

Gc

1+
s
-------

Lag

1 + s

-- --------

1 + s

G c = K -----------------1 + s
s

Lead-Lag

Gc
=

1 + 1 s

K
1 + s
------------------1

1 + 2
1+ s

1 2

------------------2

21.3 ROOT-LOCUS PLOTS

Consider the basic transform tables. A superficial examination will show that the
denominator (bottom terms) are the main factor in determining the final form of the solution. To explore this further, consider that the roots of the denominator directly impact the
partial fraction expansion and the following inverse Laplace transfer.

 When designing a controller with variable parameters (typically variable gain),

we need to determine if any of the adjustable gains will lead to an unstable system.
Root locus plots allow us to determine instabilities (poles on the right hand side
of the plane), overdamped systems (negative real roots) and oscillations (complex roots).

 Note: this procedure can take some time to do, but the results are very important
when designing a control system.
Consider the example below,

Note: This controller has adjustable gain. After this

design is built we must anticipate that all values of K
will be used. It is our responsibility to make sure that
none of the possible K values will lead to instability.
+

1
-s

1
G s =

H s = 1

-s
First, we must develop a transfer function for the entire control system.

GS s
=

K
---
s

G s
K
-----------------=
= ----- --------------------------- s + K

-------------1 + G s H s
K
1 + ---

1
s

Next, we use the characteristic equation of the denominator to find the roots
as the value of K varies. These can then be plotted on a complex plane.
Note: the value of gain K is normally found from 0 to +infinity.
s+K = 0

K
0
1
2
3
etc..

root
K

K = 0

Note: because all of the roots for all values of K are real negative this system will
always be stable, and it will always tend to have a damped response. The large the
value of K, the more stable the system becomes.

Consider the previous example, the transfer function for the whole system was
found, but then only the denominator was used to determine stability. So in general we do
not need to find the transfer function for the whole system.

Consider the general form for a negative feedback system.

G s
-- ---------
------G S= s----------1 + G s H
s

Note: two assumptions that are not often clearly

stated are that we are assuming that the control
system is a negative feedback controller, and
that when not given the feedback gain is 1.

The system response is a function of the denominator, and its roots.

1 + G s H s = 0
It is typical, (especially in textbook problems) to be given only G(s) or G(s)H(s).
The transfer function values will often be supplied in a pole zero form.
K s + z0 s + z1 s +
zm
G s H = ------------------------------------------------------------------ s + p 0 s + p 1 s + p n
s

Consider the example,

Given the system elements (you should assume negative feedback),

G s
=

-----------K
------------2
s + 3s + 2

H s = 1

First, find the characteristic equation,. and an equation for the roots,
K
1 = 0
1 + ------------ ------------ 2

s + 3s + 2
2

s + 3s + 2 + K = 0

3
1
4 K
-----9 ---
4 2 = 1.5 ----- ---- ----- --- --------------+ ------K ----- ------- -2
-2
Next, find values for the roots and plot the values,
roots =

root

0
1
2
3

0
-1
-2
-3

******CALCULATE AND PUT IN NUMBERS

21.3.1 Approximate Plotting Techniques

The basic procedure for creating root locus plots is,

1. write the characteristic equation. This includes writing the poles and zeros of
the

equation.
s + z1 s + 2z sm+ z
= 0
1 + G s H s 1 + K -------------------------------------------------------------- s + p 1 s + p 2 s + p n
=
2. count the number of poles and zeros. The difference (n-m) will indicate how
many root loci lines end at infinity (used later).
3. plot the root loci that lie on the real axis. Points will be on a root locus line if
they have an odd number of poles and zeros to the right. Draw these lines
in.
4. determine the asymptotes for the loci that go to infinity using the formula below.
Next, determine where the asymptotes intersect the real axis using the second
formula. Finally, draw the asymptotes on the graph.
180 2k + 1
k ---------------------------------=
nm

k 0 n m 1

p1 + p2 + +np 1 z +2 z + m +
z
= ------------------------------------------------------------------------------------------nm
5. the breakaway and breakin points are found next. Breakaway points exist
between two poles on the real axis. Breakin points exist between zeros. to calculate these the following polynomial must be solved. The resulting roots are
the breakin/breakout points.
A = s + p 1 s + p2 s +
pn

B = s + z 1 s + z 2 s + z m

-- --- A B A -- -- B = 0

ds

ds

6. Find the points where the loci lines intersect the imaginary axis. To do this substitute the fourier frequency for the laplace variable, and solve for the frequencies. Plot the asymptotic curves to pass through the imaginary axis at this point.
j + z1 j + z2 j m+ z
1 + K -------------------------------------------------------------------------- = 0
j + p 1 j + p 2 j + p n
Consider the example in the previous section,

Given the system elements (you should assume negative feedback),

G s
=

H s = 1

-----------K
------------2
s + 3s + 2

Step 1: (put equation in standard form)

1
= 1 + K ---------------- ----------------1+
1 + G s H s
s + 1 s + 2
K
------------ -------------
=
2 1

s + 3s + 2
Step 2: (find loci ending at infinity)
m = 0

n = 2

nm = 2

(from the poles and zeros of the previous step)

(loci end at infinity)

Step 3: (plot roots)

j

-2

-1

Step 4: (find asymptotes angles and real axis intersection)

=
k

=
0

1 8 0
k I 0 1
2----k -- +
1-- ----- ------------- -2
1 8 0
= 90
-2 --0 -- +- 1
- -- -- ---------------- --- ---2

180 2 1 + 1
=
= 270
-----------------------------------2
0 1
2
=
----------------------------- =
0
2

-2

-1

asymptotes

linear feedback controller analysis - 21.20

Step 5: (find the breakout points for the roots)

B = s + 3s + 2

d
-- --- A =

d
-- ---

0
ds

B =
ds

2s + 3

A -- -- B B -- --- A = 0

ds

A = 1

ds

1 2s + 3 s + 3s + 2 0 =
0

-2

-1

-1.5

2s + 3 = 0
s = 1.5
Note: because the loci do not intersect the imaginary axis, we know the system will be
stable, so step 6 is not necessary, but we it will be done for illustrative purposes.

Step 6: (find the imaginary intercepts)

1 + G s H s = 0
1+K
1
------------ ------------- =
0
2
s + 3s + 2
2

s + 3s + 2 + K = 0
2

j + 3 j + 2 + K = 0
2

+ 3j + 2 + K = 0
2

+ 3j + 2 K = 0

3j 4K 1
=
=
= ------------------------------2
------------------------------------------ --------------------------------------------------------------2
2
3j 3j 2 4 2 K

3j 9 + 8 + 4K

In this case the frequency has an imaginary value. This means that there will be
no frequency that will intercept the imaginary axis.

linear feedback controller analysis - 21.21

Plot the root locus diagram for the function below,

--s
G
K + --- 5
= s H s ------ - ------ ----------2
s s + 4s + 8

21.4 DESIGN OF CONTINUOUS CONTROLLERS

21.5 SUMMARY

21.6 PRACTICE PROBLEMS

3. Given the transfer function below, and the input x(s), find the output y(t) as a function
of time.
y s
-------- s = ----x
5
-----s+2

x t = 5

t 0 sec

8. Draw a detailed root locus diagram for the transfer function below. Be careful to specify
angles of departure, ranges for breakout/breakin points, and gains and frequency at stability
limits.
2

2K s + 0.5 s + 2s + 2
G s = ---------------------------------------------------------3
s s + 1 s +
2
10. Draw the root locus diagram for the transfer function below,
2
K s + 4
--- --------- G
=s
-------2
s s + 1
11. Draw the root locus diagram for the transfer function below,
K s + 1 s + 2
G s = -----------------------------------3
s
12. The block diagram below is for a motor position control system. The system has a proportional controller with a variable gain K.
d

Vd +

2
-

Ve
K

Vs
1 0 0
-- --- --- --s+2

Va
2
a) Simplify the block diagram to a single transfer function.

ans.

2 0 0 K
-------------- --- ---- -------------2
s + 2s + 200K

1
-s

b)
D

raw the Root-Locus diagram for the system (as K varies). Use either the
approximate or exact techniques.

ans.

roots =
= 1 1 200K
2
-----4 --
4 2
-----0 0 ----K ---- - -- ---- ---- ---------2
Im
K
r
o
K=0.005
o
ts
0
-2
-1

Re

0
0

c) Select a K value that will result in an overall damping coefficient of 1. State if the Root-Locus
diagram shows that the system is stable for the chosen K.

2
(ans. s 2 + 2s + 200K = 2
s + 2 n s + n

K = 0.005

n = 1

From the root locus graph this value is critically stable.

13. Draw a Bode Plot for either one of the two transfer functions below.
s + 1 s +
1000
----------------------------------------2
s + 100

OR

5
- --2
s

15. Given the system transfer function below.

20K
o = ----------------------------2
----s + s + 20K
d
a) Draw the root locus diagram and state what values of K are acceptable.
b) Select a gain value for K that has either a damping factor of 0.707 or a natural
frequency of 3 rad/sec.
c) Given a gain of K=10 find the steady state response to an input step of 1 rad.
d) Given a gain of K=10 find the response of the system as
17. The equation below describes a dynamic system. The input is F and the output is V. It has
the initial values specified. The following questions ask you to find the system response to a

unit step input using various techniques.

V 0 =
V' 0 = 2
V'' + 10V' + 20V = 20F
1
a) Find the response using Laplace transforms.
b) Find the response using the homogenous and particular solutions.
c) Put the equation is state variable form, and solve it using your calculator.
Sketch the result accurately below.
18. A feedback control system is shown below. The system incorporates a PID controller. The
closed loop transfer function is given.
X +
-

----K +K
+ Kd s
s
p
i

3
----- -----s+9

Y
- -- =
---------X

s 2 3K d + s 3K p + 3K i
----------

---------------------------------------------------------------2
+ 3K i
s 4 + 3K d + s 36 +
3K p

a) Verify the close loop controller function given.

b) Draw a root locus plot for the controller if Kp=1 and Ki=1. Identify any values
of Kd that would leave the system unstable.
c) Draw a Bode plot for the feedback system if Kd=Kp=Ki=1.
d) Select controller values that will result in a natural frequency of 2 rad/sec and
damping coefficient of 0.5. Verify that the controller will be stable.
e) For the parameters found in the last step find the initial and final values.
f) If the values of Kd=1 and Ki=Kd=0, find the response to a ramp input as a function of time.
19. The following system is a feedback controller for an elevator. It uses a desired heigh d provided by a user, and the actual height of the elevator h. The difference between these two is
called the error e. The PID controller will examine the value e and then control the speed
of the lift motor with a control voltage c. The elevator and controller are described with
transfer functions, as shown below. all of these equations can be combined into a single
system transfer

equation as shown.
e = dh

error

h
2

Ki
+ ----+K
d s
=
s
p

2 s
- --- ---+
1
----- ---+ s
----- --s
combine the transfer functions
-- =
K
e

PID controller

-c

1 0
= ---- --- -----2

elevator

s +s

2
h
c h--
10 s + 1
2s + 1 + s
s +
1

1 0--- --- 0 - = ---------------------= -= --------1 ----

-----------=
---------------2
2
s s + 1
e c e
s
s +s
----- ------ -----s
s
=
1 0 s + 1
h
--------- --- -------- ----eliminate e
-----2
dh
s
1 ---0 --s +
1
h = ------ ----- - d h
2
s
10 s + 1
10 s + 1 d
h 1 + --------------=
--------------------------

2
2
s
s
0 s + 1
1----------- ----

-- =
s
-------------- -----------------

d
1 0 s
-- 1 + - -+
1
----- ------- 2 ---

1 0
= ------ --- --s + 1
---- ----- -0
-------

system transfer function

s + 10s + 10

a) Find the response of the final equation to a step input. The system starts at rest
on the ground floor, and the input (desired height) changes to 20 as a step input.
b) Write find the damping coefficient and natural frequency of the results in part
a).
c) verify the solution using the initial and final value theorems.

(ans. a)

d s
=

d t = 20u t

1 0 s
= ------ --- --- ---h +-----1--0-------2
d
s + 10s + 10

2 0
- -- -s

C
20

1 0 s
h = ------ -- ---- --- = --- + -------------------------------- + --------------------------------+
1 0
----- --2 ------- ----s s + 5 3.873 j s + 5 + 3.873j
s + 10s + 10 s
200s + 200 = 20
A = lim

------------------------------
s0 2
s + 10s + 10
-------2---0---0---- = 2.5 22.6j
B =
lim
s
+ 2 0
---- ---- ---- --0
-------
s 5 + 3.873 j s s + 5 + 3.873j
C = 2.5 + 22.6j

h t = 20 + L

A =

2 . 5
2 2 .
-- ---- -- -- ---- ----- --- --- 6 j
2 . 5 +
---- - + -- ---- -- -- ---- ----2 2 . 6 j
--- --- -- --- -s + 5 3.873j s + 5 + 3.873j

= 4.602
= 5
= 3.873

h t

= 20 + 2 22.73 e
4.602

b)

5 t

= n

cos 3.873t +
5
------=n
2 5
1 - 2--- n-
n

3.873 =

1 n =

2
25 2
25
1 ----- =

15 =

c)

h 0 =

lim
s

n =

35 = 5.916

= = 0.845
---- ----35
5

1 0 s
s ------ -- ---- --+ 1 0
20
----- --- -------
2
s + 10s + 10 s

h =

= lim 1 0 s
- -- ----- 20 = 0
s

2.5 + 22.6 = 22.73

lim

1 0 s +
1
--- --- ---- ----- --s
0
2 0
------ ------- -- -2
s0
s + 10s + 10 s

10 20
= ---------------- = 20
10

21.7 PRACTICE PROBLEM SOLUTIONS

21.8 ASSIGNMENT PROBLEMS