Dynamics and Control II

MIT OpenCourseWare
http://ocw.mit.edu
2.004 Dynamics and Control II
Spring 2008
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
R
C
v(t)
+
-
V (t)
in
Usi ng KVL & KCL:
RC + v = V (t)
dv
dt
in
F(t)
force
m
mass
B (vi scous fri cti on)
v(t)
vel oci ty
From a force bal ance:
+ v = F (t)
m dv
B dt
1
B
f
s
P
C
R
g
f
f
fl ui d resi stance
fl ui d capaci tance
p
pressure at
bottom of tank
pump pressure
output fl ow
Usi ng fl ui d j uncti on equati ons
R C + p = P (t)
dp
dt
f f

Massachusetts Institute of Technology
Department of Mechanical Engineering
Spring Term 2008
Lecture 1
1
Reading:
Nise: Chapter 1
1 Elements of the Course
(a) System Dynamics: System dynamics provides a unied approach to the modeling
and dynamic behavior of linear systems in many energy domains. For example:
a) An electrical network:
b) A simple mechanical system:
c) A uidic system:
1
copyright c D.Rowell 2008
11
Li near system
K(J) O(J)
output
i nput
K(J)
J
System i s usual l y represented by an ODE
wi th constant coeffi ci ents.
y(J)
J
T
+
-
v
i
9
i = - T
1
K
__
v
v = K 9
v
el ectri cal mechani cal (rotati onal )
angul ar vel oci ty
torque
current
vol tage

We recognize a common form to the ODE describing each system and create analogs
in the various energy domains, for example
voltage current electrical
velocity
pressure
and
force
volume ow rate
in the
mechanical
uidic
domains.
Then for each system, we can write a dierential equation:
dy
+ y = u(t)
dt
where u(t) is the system input
y(t) is the system output, and
is a system parameter (time-constant).
We will frequently use block diagrams to represent the input/output relationships of
systems
usually represented by an ODE with constant coecients.
In addition we will investigate energy transducers that convert power/energy from
one domain to another, for example a motor converts power P from the electrical do
main to the rotational domain.
A transducer is bi-directional, that is it can transmit power P in both directional, so
that the motor can also act as a generator.
transduction
P = vi motor P = T
generator
12
electrical rotational
5
u(t) y(t)
output
i nput
+ 2 z w + w = u
d y
dt
2
2
dy
dt
n
n
2
system represented i n a "standard" form
"Open-l oop" system
K(J)
O(J)
output
i nput
Control l er
Actuator
System
Sensor
desi red response
actual response
measurement
feedback path
(motor, etc)
(b) Linear System Theory A generalized method for the description of the dynamic
response of systems described by linear ordinary dierential equations (ODEs) with
constant coecients without regard to the particular energy domain.
Typical questions we might ask about the system might be what is the response to:
(a) a sinusoidal input?
(b) a step input ?
(c) a short pulse ?
(c) Feedback Control Theory The use of feedback to modify the dynamic behavior
of a system
Often the inherent system behavior is unsatisfactory, for example:
The response might be too slow.
The response might be unstable.
The system might be susceptible to external inuences.
Components inside the system might change their values as they age, causing
the system response to change.
Often control involves monitoring the system response - comparing the response to the
desired behavior - and generating a system input so as to drive the system toward the
desired response.
13
furnace
thermostat
i nternal temperature 6
heat from furnace
external temperature 6
heat l oss from house
o
6
6
@
6 + ,
@
6 - , @
Furnace heat
on
off
J
J
(1) monitor the response in real-time.
(2) compare the actual response with the desired response.
(3) decide how to modify the response by changing the input.
Example 1
Temperature control in a home:
Desired behavior - maintain temperature at T
Control law (algorithm) using a thermostat:
(a) if T > T
0
+ - turn o furnace.
(b) if T < T
0
- turn on furnace.
(c) if T
0
T T
0
+ - do nothing.
A typical response as the thermostat turns on and o might be:
Notice that the temperature rises while the furnace is on, and falls while it is
o. The thermostat acts to keep the temperature centered about the desired
value T
d
(known as the set-point).
14
F (t)
F
d
p
F (t)
m
mass
v(t)
vel oci ty
F (t)
d
p
v(t)
engi ne
K
e
m + Bv = F
dv
dt
+
+
F (t)
F (t)
d
p
G
gas pedal
car model
V
car speed
Example 2
Cruise Control for a car:
Goals - maintain the speed of a car at a prescribed value in the presence
of external disturbances (external forces such as wind gusts, gravita
tional forces on a incline, etc).
Also - improve the dynamic response of the car as the driver steps on the
gas.
(a) Form a dynamic model of the car. Assume a simplied model
Model the car as a simple lumped mass element m, sliding on a
viscous friction element, F
B
= BV
B
- simplication
Assume two external forces
F
p
(t) - the propulsive force F
p
(t) from the engine.
F
d
(t) - a disturbance force F
d
(t) from the environment.
Also assume F
p
(t) = K
e
(t) where (t) is the gas-pedal depression
and K
e
is a constant. Then from a simple force balance:
dv
m + Bv = F
p
(t) + F
d
(t)
dt
dv
m + Bv = K
e
(t) + F
d
(t)
dt
and draw a block diagram
(b) Closed-loop Control:
Now design the controller. Assume we will use error-based control. In
other words, given a desired speed v
d
, and the measured car speed v(t),
we dene the error e(t) as
e(t) = v
d
(t) v(t)
15
engi ne
K
e
m + Bv = F
dv
dt
+
+
F (t)
F (t)
d
p
G
car model
V
car speed
K
c
e(t) +
feedback path
V (t)
d
-
error
control l er
and choose a control law that tells us to depress the gas pedal by an
amount proportional to the error:
(t) = K
c
e(t) = K
c
(v
d
(t) v(t)),
so that the propulsive force acting on the car is
F
p
(t) = K
e
K
p
e(t) = K
e
K
c
(v
d
(t) v(t)),
This is known as proportional control.
16
F (t)
F
d
p
F (t)
m
mass
v(t)
vel oci ty
F (t)
d
p
v(t)
engi ne
K
e
m + Bv = F
dv
dt
+
+
F (t)
F (t)
d
p
G
gas pedal
car model
V
car speed

Spring Term 2008
Lecture 2
1
Reading:
Nise: Chapter 1
1 Cruise Control Example (continued from Lecture 1)
From Lecture 1 the model for the car is:
where the propulsion force F
p
(t) is proportional to the gas pedal depression :
F
p
(t) = K(t)
so that
dv
m + Bv = K
e
(t) + F
d
(t)
dt
The Open-LoopDynamic Response: Lets examine how the car will respond to
commands at the gas-pedal. Assume F
d
(t) = 0, so that we can write the dierential equation
as
m K
e
v + v =
B B
and compare this to the standard form for a rst-order ODE:
y + y = f(t)
where the time-constant = m/B, and the forcing function f(t) =
K
B
e
(t).
1
21
0
0
J
J !J "J t (secs)
v
o
0. 6v
o
0. 8v
o
0. 4v
o
0. 2vo
vel oci ty v(t)
e = 0. 3679
e = 0. 1 353
e = 0. 0498
e = 0. 01 83
-1
-2
-3
-4
Consider
1) The coast-down response from an initial speed v(0) = v
0
with = 0. The homogeneous
response is
v(t) = v
0
e
= v
0
e
t
B
m
t
which has the form

and after a period 4 we have v < 0.02v
0
.
2) The response to a step in the command (t). Assume that the car is at rest, that is
v(0) = 0, and we step on the gas so that (t) = . The dierential equation is then
m K
e
v + v =
B B
with v(0) = 0.
(a) The steady-state (nal) speed is
K
e
v
ss
=
B
(found by letting all derivatives to be zero), and
(b) the dierential equation may be solved to give the dynamic response:
B
m
t K
e

v(t) = v
ss
(1 e

) = (1 e
t
)
B
which is shown below:
22
engi ne
K
e
m + Bv = F
dv
dt
+
+
F (t)
F (t)
d
p
G
car model
V
car speed
K
c
e(t) +
feedback path
V (t)
d
-
error
control l er
0
J
L
II
v(J)
4J
After 4J the response i s wi thi n
2% of the fi nal val ue.
steady-state
regi on
transi ent regi on
2 Closed-Loop Control
Now design the controller. Assume we will use error-based control. In other words, given
a desired speed for the car v
d
, and the measured car speed v(t), we dene the error
e(t) as the dierence
e(t) = v
d
(t) v(t),
and choose a control law that is based on e(t)
(t) = K
c
e(t) = K
c
(v
d
(t) v(t))
where K
c
is the controller gain. Thus the control eort is proportional to the error, and
the block diagram becomes:
Note that the control law states:
if v < v
d
then e > 0 - depress gas pedal.
if v = v
d
then e = 0 - set = 0 (do nothing).
if v > v
d
then e < 0 - set < 0 (apply brakes).
The new dierential equation is
dv
m + Bv = K
c
K
e
(v
d
(t) v) + F
d
(t)
dt
and rearranging
dv
m + (B + K
c
K
e
)v = K
c
K
e
v
d
(t) + F
d
(t)
dt
23
steady-state transi ent
t
y = K
ss
y(t)
4t
at t = 4t, the response i s approx.
0. 98y .
ss
0
which is the closed-loop dierential equation. We can write this as
m dv K
c
K
e
1
+ v = v
d
(t) + F
d
(t)
B + K
c
K
e
dt B + K
c
K
e
B + K
c
K
e
and compare with the standard rst-order form
y + y = u(t)
where = m/(B + K
c
K
e
) is the closed-loop time-constant.
The important thing to note is that feedback has modied the ODE.
3 Questions:
Assume there are no external disturbance forces, that is F
d
(t) 0 for now,
(a) If we command the car to travel at a steady speed v
d
, what speed will it actually reach?
To nd the steady-state speed v
ss
, set
dv
= 0 and solve for v
ss
, giving
dt
K
c
K
e
v
ss
= v
d
B + K
c
K
e
or v
ss
< v
d
, for B > 0.
Note that as we increase the controller gain so that K
c
K
e
B then v
ss
v
d
.
(b) How has the dynamic response aected by the feedback?
Consider the rst-order ODE
dy
+ y = Ku(t).
dt
The response to a steady input u(t) = 1 with initial condition y(0) = 0 is
t
y(t) = K(1 e

)
24
K i ncreasi ng
steady-state speed
approaches v .
t
c
d
v
v
d
J decreases
0
F (t)
p
v(t)
F
(t)
F
=
m
g
s
i n
(
G
d
p
v
(t)
G
(a) on l evel ground
(b) on an i ncl i ne
For the car under feedback control, this translates to a steady-state speed
K
c
K
e
v
ss
= v
d
B + K
c
K
e
which is a function of the controller gain K
c
, and a closed-loop time constant that is
m
=
B + K
c
K
e
which is also a function of K
c
, and as K
c
increases, decreases, meaning that the car
responds more quickly to changes in the gas pedal.
4 The Eect of an External Disturbance Force
Now consider the car on an incline under closed-loop control:
(a) On horizontal ground F
d
= 0, and we showed:
K
c
K
e
v
ss
= v
d
B + K
c
K
e
(b) On the incline there is a constant disturbance force F
d
= mg sin acting down the
incline, and from the closed-loop dierential equation:
m dv K
c
K
e
1
+ v = v
d
(t) + F
d
(t)
B + K
c
K
e
dt B + K
c
K
e
B + K
c
K
e
the steady-state speed will be
K
c
K
e
mg sin
v
ss
= ,
B + K
c
K
e
v
d

B + K
p
K
e
25
but we note that as the controller gain K
p
is increased the impact of the disturbance
is reduced.
Conclusion: Feedback control reduces the eect of external disturbances on the
system behavior.
26
System descri bed by
"transfer functi on"
ODE
u(t)
U(s)
Y(s)
y(t)
ti me domai n
Lapl ace domai n
sol ve the ODE
sol ve by mul ti pl i cati on
L L
-1
transform system
representati on

Spring Term 2008
Lecture 3
1
Reading:
Nise: Secs. 2.2 and 2.3 (pp. 32 - 45)
1 The Laplace Transform
The reason that the Laplace transform is useful to us in 2.004 that it allows algebraic
manipulation of ordinary dierential equations
1) Solution of ODEs is dicult, so
2) Transform the problem to a domain where the solution is easier.
3) Solve the problem in the new domain.
4) Perform the inverse transform to move the solution back to the original domain (if we
need to).
Denition: The one-sided Laplace transform is an integral transform dened as

F (s) = L{f(t)} = f(t)e

st
dt
0

It maps the function f(t) to a function of the complex variable s = + j (j =
1), and
F (s) is itself generally complex. We also write
f(t) = L
1
{F (s)}
1
31
K J
I
BJ
J
/6
6 J
area = x T = 1
6
B(J)
for the inverse transform and often
f(t)
L
F (s)
1.1 Some Common Examples Used in Control Theory:
(a) The unit step function: u
s
(t):

1
[e
st
]
1
U
s
(s) =
0

1e
st
dt =
s
0
=
s
(b) The one-sided exponential f(t) = u
s
(t)e
at
(a > 0)
1

e
at
e
st
dt = F (s) =
0
s + a
(c) A very brief with unit area:
1

T
F (s)
T
0

1dt = 1
As T 0, the amplitude becomes very large and we dene the Dirac delta (or impulse)
function (t) as:
32
(t) = 0 for all t = 0
(t) is undened (innite) at t = 0

(t)dt = 1 (unit area).
and L{(t)} = 1 from above.

1.2 The Inverse Laplace Transform
If F (s) = L{f(t)}, then f(t) = L
1
{F (s)} where L
1
denotes the inverse transform. In
general L
1
{} requires integration along a contour in the complex s = + j plane, parallel
to the imaginary axis. This is rarely done in practice.
Instead, break up F (s) into a sum of functions with known L
1
{}, and use table lookup,
for example:
Example 1
Find the inverse Laplace transform of
4
F (s) = .
s
2
+ 5s + 6
4 4 4 4
F (s) =
s
2
+ 5s + 6
=
(s + 3)(s + 2)
=
s + 2

s + 3
(partial fractions)
and since Le
at
=
1
, we recognize
s+a

4

4
f(t) = L
1
{F (s)} = L
1
s + 2
L
1
s + 3
= 4e
2t
4e
3t
Note: See Nise for treatment of repeated roots
1.3 Properties of the Laplace transfrom:
We will discuss only the major properties that are useful in 2.004:
(a) Linearity: If F (s) = L{f(t)} and G(s) = L{g(t)} then
L{af(t) + bg(t)} = aF (s) + bG(s)
where a and b are constants
The linearity property is fundamental to our treatment of ODEs and linear systems.
33
ft
t
ft-J
t
J
0
(b) Time Shift: If F (s) = L{f(t)} then
L{f(t )} = e
s
F (s)
This is an important property in control theory because pure delays aect system
stability under feedback control.
(c) Dierentiation Property: If F (s) = L{f(t)}, then
df
L{
dt
} = sF (s) f(o
)
d
2
f
L{
dt
2
} = s
2
F (s) sf(o
) f
(0)
and
n
L{
d
dt
n
n
f
} = s
n
F (s)

s
nk
f
(k1)
(0)
k=1
This is perhaps the most important property in this course.
(d) Integration property: If F (s) = L{f(t)},
t

1
L{
0
f()d} =
s
F (s)
(e) The Final Value Theorem: If F (s) = L{f(t)}, then
lim f(t) = lim sF (s)
t s0
provided the limit exists. The f.v. theorem is useful for determining the steady-state
response of systems.
2 Laplace Domain System Representation
Suppose that through modeling we have found that a system is described by a dierential
equation
d
n
y d
n1
y dy d
m
u d
m1
u du
a
n
dt
n
+ a
n1
dt
n1
+ . . . + a
1
dt
+ a
0
y = b
m
dt
m
+ b
m1
dt
m1
+ . . . + b
1
dt
+ b
0
u
34
Transfer functi on
0(I)
K(J)
;(I) = 7(I)0(I)
O(J)
ti me domai n
Lapl ace domai n
L
L
-1
7(I)
mul ti pl i cati on
ODE
ti me domai n
Lapl ace domai n
sol ve the ODE
L
L
-1
?
what i s the process?
U(s)
u(t)
Y(s)
y(t)
Assume that the system is at rest at time t = 0, that is y(0) = 0, y(0) = 0, etc. and that
u(0
) = 0, u(0
) = 0, etc then using the dierentiation property of the Laplace transform

on each term in the ODE gives:
a
n
s
n
Y (s) + a
n1
s
n1
Y (s) + . . . + a
0
Y (s) = b
m
s
m
U(s) + b
m1
s
m1
U(s) + . . . + b
0
U(s)
[a
n
s
n
+ a
n1
s
m
+ b
m1
s
n1
+ . . . + a
0
]Y (s) = [b
m
s
m1
+ . . . + b
0
]U(s)
and solving for Y (s)
b
m
s
m
+ b
m1
s
m1
+ . . . + b
0
Y (s) = U(s) = H(s)U(s)
a
n
s
n
+ a
n1
s
n1
+ . . . + a
0
where H(s) is dened as the system transfer function.
b
m
s
m
+ b
m1
s
m1
+ . . . + b
0
N(s)
H(s) = =
a
n
s
n
+ a
n1
s
n1
+ . . . + a
0
D(s)
where numerator coecients come from the RHS of the ODE and the denominator coe
cients come from the LHS. H(s) is a rational fraction for most linear systems.
The Laplace transform (transfer function) has changed the system represen
tation to from an ODE to an algebraic representation with a multiplicative
input/output relationship.
In system dynamics and control work we use the transfer function as the primary system
representation.
35
F (t)
m
mass
vel oci ty
v(t)
m + Bv = F(t)
dv
dt
Example 2
Find the transfer function of a system represented by the ODE:
d
3
y d
2
y dy du
5 + 17 + 6 + 5y = 8 + 6u
dt
3
dt
2
dt dt
Answer:
8s + 6
H(s) =
5s
3
+ 17s
2
+ 6s + 5
Note: A fundamental assumption when using the transfer function to compute
responses is that the system is at rest at time t = 0.
Example 3
Find the response V (s) of the velocity of the mass element shown below to a unit
step the applied force F (t)
From the dierential equation
(ms + B)V (s) = F (s)
1
V (s) = F (s).
ms + B
For a unit-step in the force F (t) , F (s) = 1/s and
1 1 B B
V (s) =
ms + B

s
=
s

s + B/m
Taking the inverse Laplace transform gives the response
B
m
v(t) = L
1
{V (s)} =
1
1 e
t

B
36
R
C
v t
o
V t
in
i
+
-

Example 4
Find the transfer function relating: a) x(t), b) v(t) to F (t) for the system
F (t)
m
mass
spri ng
K
B
damper
v(t)
x(t)
a) From a force balance
mx + Bx + Kx = F.
By inspection
X(s) 1
H(s) = =
F (s) ms
2
+ Bs + k
b) Since v(t) = dx/dt,
t
mv + Bv + K vdt = F
0
or
mv + Bv + Kv = F
and
V (s) s
H(s) = =
F (s) ms
2
+ Bs + K
Example 5
Find the transfer function of the electrical circuit
What we know:
(1)
V
in
(t) = v
c
+ v
r
(KVL)
37
(2)
v
o
= v
R
=
1
R
i
R
(3)
i
R
= i
c
(KCL)
From (1):
t
1
V
in
(t) = v
c
+ v
R
= i
c
dt + v
R
C
0
Dierentiate and use (2):
1 1
V
in
(t) = i
C
+ v
R
= v
R
+ v
R
C RC
Use v
o
= v
R
to obtain:
V
in
(t) =
1
v
o
+ v
o
.
RC
Take the Laplace transform of both sides, and use the derivative property to give
V
o
(s) RCs
H(s) = =
V
in
(s) RCs + 1
38
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
0 I0 I

0 I 0 I

7I
;I 7I ;I
:I
H s
1
2
+
+
H (s + H I
1 2
Ys Ys Us
Us
H s

Spring Term 2008
Lecture 4
1
Reading:
Nise: Secs. 5.15.3
1 Block Diagram Algebra (Interconnection Rules)
a)Series (Cascade) Connection:
Since the output of the rst block is X(s) = H
1
(s)U(s),
Y (s) = H
2
(s)X(s) = H
1
(s)H
2
(s)U(s)
Note: This is only true if the connection of H
s
(s) to H
1
(s) does not alter the output of
H
1
(s) known as the non-loading condition.
b)Parallel Connection In this case the input U(s) is applied to both inputs and the
outputs are summed:
Y (s) = H
1
(s)U(s) + H
2
(s)U(s) = (H
1
(s) + H
2
(s))U(s)
Example 1
Express
6
H(s) =
s
2
+ 5s + 6
1
41
7(I)
;(I)
3
I + 3
2
I + 2
order of bl ocks i s arbi trary
-
+
Y(s)
U(s)
6
s + 2
6
s + 3
+
Ys
U s
1
U s
2
Hs
-
+
Ys
U s
1
U s
2
Hs
Hs
+
7I
;I 0 I
0 I
7I
;I 0 I
0 I
as (a) a series connection, and (b) a parallel connection of rst-order blocks

a) Series:
6 3 2
H(s) = =
s
2
+ 5s + 6 s + 3

s + 2
b) Parallel: Using partial fractions we nd
6 6
H(s) =
s + 2

s + 3
Notes:
a) These two systems are equivalent.
b) A partial fraction expansion is eectively a parallel implementation.
c) A factored representation of H(s) is eectively a series implementation.
c)Associative Rule:
Y (s) = (U
1
(s) + U
2
(s))H(s) U
1
(s)H(s) + U
2
(s)H(s)
d)Commutative Rule:
The order does not matter in a series connection.
42
G (s)
c
G (s)
p
E(s)
R(s) C(s)
control l er pl ant
error
+
-
feedback path
reference
i nput
control l ed
output
G (s)
c
G (s)
p
E(s)
V(s)
control l er car dynami cs
error
+
-
feedback path
desi red
speed
actual
speed
V (s)
d
G(s)
2 The Closed-Loop Transfer Function
a) Unity feedback
Notes:
(a) The term unity feedback means that the actual output value is used to generate the
error signal (the feedback gain is 1).
(b) In control theory transfer functions in the forward path are often designated by G(s)
(see below).
(c) It is common to use R(s) to designate the reference (desired) input, and C(s) to desig
nate the controlled (output) variable.
From the block diagram:
C(s) = (G
p
(s)G
c
(s))E(s)
and
E(s) = R(s) C(s)
or
C(s) = G
p
(s)G
c
(s)(R(s) C(s)
Rearranging:
C(s) G
c
(s)G
p
s
G
cl
(s) = =
R(s) 1 + G
c
(s)G
p
s
is the unity feedback closed-loop transfer function.
Example 2
Find the closed-loop transfer function for the automobile cruise control example:
43
For the car mv + Bv = F
p
= K
e
so that
V (s) K
s
G
p
(s) = =
(s) ms + B
For the controller (s) = K
c
E(s)
(s)
G
c
(s) = = K
c
E(s)
Then from above
V (s) G
c
(s)G
p
(s)
G
cl
(s) = =
V
d
(s) 1 + G
c
(s)G
p
(s)
K
c
K
e
ms+B
K
c
K
e
G
cl
(s) = = ,
1 +
K
c
K
e
ms + (B + K
c
K
e
)
ms+B
and by inspection the closed-loop dierential equation is
mv + (B + K
c
K
e
)v = K
c
K
e
v
d
.
Aside:
Use the Laplace transform nal value theorem to nd the steady state velocity to a
step input v
d
(t) = v
d
For the step input
v
d
v
d
(s) =
s
and in the Laplace domain
K
c
K
e
v
d
v(s) = G
c1
(s)V
d
(s) =
ms + (B + K
c
K
e
) s
The F.V. theorem states lim
t
f(t) = lim
s 0
sF (s) so that
K
c
K
e
v
d
v
ss
= lim
t
v(t) = lim
s0
s
ms + (B + K
c
K
e
) s
K
c
K
e
v
s
s =
B + K
c
K
e
which is same as we obtained before.
3 Closed-Loop Transfer Function With Sensor Dynamics:
Until now we have assumed that the output variable y(t) is measured instantaneously, and
without error. Frequently the sensor has its own dynamics - for example the sensor might
be temperature measuring device modeled as a rst-order system:
44
4
+
from sensor
to control l er
8
sensor
8
out
fi l ter
1
J I + 1
I
system
output
sensor
feedback
si gnal
J
J
where
s
is the sensor time constant.
The closed-loop block diagram is
G (s)
c
G (s)
p
E(s)
R(s) C(s)
control l er pl ant
error
+
- reference
i nput
actual
output
H(s)
sensor
i ndi cated output
where H(s) is the transfer function of the sensor. In this case:
C(s) = (G
c
(s)G
p
(s))E(s)
but now E(s) is the indicated error (as opposed to the actual error):
E(s) = R(s) H(s)C(s)
so
C(s) = G
c
(s)G
p
(s)(R(s) H(s)C(s))
or
C(s)(1 + G
e
(s)G
p
(s)H(s)) = G
e
(s)G
p
(s)H(s).
C(s) G
c
(s)G
p
(s)
G
cl
(s) = =
R(s) 1 + G
c
(s)G
p
(s)H(s)
is the modied closed-loop transfer function.
Example 3
Suppose that velocity sensor in the cruise control is noisy, and a simple elec
trical lter is used to smooth the output. Find the eect of the lter on the
closed-loop dynamics.
45
c
E(s)
V (s)
control l er pl ant
error
+
-
reference
i nput
actual
speed
fi l ter
i ndi cated speed
K
K
ms + B
e
1
RCs + 1
d
V(s)
Using Kirchos Voltage Law (KVL) we nd
RC v
out
+ v
out
= v
sensor
so that
V
out
(s) 1
H(s) = =
V
sensor
(s) RCs + 1
Then the closed-loop transfer function is
K
c
K
e
V (s) G
c
(s)G
p
(s)
ms+B
= = = G
cl
(s)
V
a
(s) 1 + G
c
(s)G
p
(s)H(s) 1 +
K
c
K
e
(ms+B)(RCs+1)
K
c
K
e
(RCs + 1)
=
(ms + B)(RSs + 1) + K
c
K
e
K
c
K
e
(RCs + 1)
G
cl
(s) =
mRCs
2
+ (BRC + m)s + (B + K
c
K
e
)
and the dierential equation relating the speed of the car to the desired speed
command is now
mRCv + (BRC + m) v + (B + K
c
K
e
) v = K
c
K
e
RC v
d
+ K
c
K
e
v
d
and we note:
1) we now have a second-order system - the dynamics may change signicantly,
2) we have derivative action on the RHS of the dierential equation.
46
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
i
The coi l of wi re has properti es
of inductance (energy storage i n
the magneti c fi el d) and resistance
(energy di ssi pati on).

Spring Term 2008
Lecture 5
1
Reading:
Class Handout: Modeling Part 1: Energy and Power Flow in Linear Systems
Sec. 1 (Introduction)
Sec. 4 (Electrical System Elements)
1 Modeling of Physical Systems
We will develop a unied modeling method that will allow the generation of a transfer
function in several energy domains. The method is based on an extension of impedance
concepts used in electrical systems. For that reason we will start with electrical systems.
Lumped parameter modeling involves approximating physical elements (where the param
eters may be distributed in space) with discrete elements that are assumed to be concentrated
at poins (or lumped) in space. For example, a real electrical inductor is made from a coil
of wire.
The coil has the primary property of electrical inductance, BUT it also has property of
resistance associated with the wire - this resistance is distributed throughout the coil. In
lumped parameter modeling we approximate the coil as having two elements - an inductance
L and a lumped resistance R.
2 Modeling Electrical Systems
Step 1: Identify the variables to be used. We will concentrate on the use of power
variables, that is a pair of variables whose product is power. In electrical systems:
P = vi
where P is power, v is the voltage drop (volts) across an element, and i is the current
(amps) through an element. We therefore use v and i as our modeling variables.
1
51
Area A
d
Di el ectri c
permi ti vi ty A
+
-
C =
A A
d
___
+ + + + + + + + + + + +
- - - - - -
- - - - - -
el ectri c fi el d
E
i
+
-
Step 2: Identify the primitive lumped modeling elements. In each energy do
main we will identify 3 elements, two elements that store energy, and an element that
dissipates energy. In the electrical domain the energy storage elements are the capacitor
and the inductor, and the dissipative element is the resistor.
The Capacitor: The capacitor stores energy in the electrostatic eld between its
plates.
For a capacitor the current and voltage drop are related by
dv
i = C
dt
where C is dened as the capacitance (Farad), or alternatively
t
1

v = idt + v(0)
C
0
Also, if we dene charge as q =
0
t
idt then
1
v = q
C
For a parallel plate capacitor,where the plates have area A, and are separated by
a distance d
A A
C
d
= C =
d
where is dened to be the permittivity of the (non-conducting) dielectric material
between the plates.
Note that power can ow in and out of a capacitor, since
dv
P = vi = Cv
dt
and v and dv/dt may have same or opposite signs. The capacitor is therefore an
energy storage element. The energy stored in a capacitor is
t

1
E
C
(t) = Pdt + E(0) = Cv
2
.
2
0
52
L
v v
1 2
1 2
i
v = v - v
2 1
N turns on l ength l
Core of area A
and permeabi l i ty m
L =
mN A
l
_____
2
Hel i cal l y wound coi l
Notes:
1) The Farad is a very large unit. Most practical capacitors are mea
sured in units of
microfarads ( F) = 10
6
F,
nanofarads (nF) = 10
9
F, or
picofarads (pF) = 10
12
F
2) A capacitor will not pass a dc current. Current only ows when
the applied voltage is changing.
The Inductor:
Inductance is a property that represents energy stored in the electromagnetic eld
in a current carrying conductor. A practical inductor consists of a coil of wire
Lenz Law relates the voltage across an inductor to yje current owing through
it:
di
v = L
dt
or
t
1
i = vdt + i(0)
L
0
where L is the inductance, with units of the Henry. Power can ow in and out of
an inductor since
di
P = vi = Li ,
dt
and i and di/dt may have the same or opposite signs. The stored energy is
t

1
E = Pdt + E(0) = Li
2
2
0
53
L
R
v = v - v
2 1
2 1
i
R
v
2
v
1
i
+
-
C
i
i
+
-
L
-
+
R
i = C
dv
dt
v = L
di
dt
v = Ri
Notes:
1) The Henry is a very large unit. Practical inductors have values in
millihenrys (mH) - 10
3
H
microhenrys (H) - 10
6
H
2) Notice that a pure inductance will have v = 0 if
di
= 0 (a dc current)
dt
3) Practical inductors are wound with (copper) wire and will have a nite
resistance R. The lumped model including both phenomena is
The relative importance of the resistance is dependent on the operating con
ditions.
The Resistor:
The current and voltage in a resistor are governed by Ohms law:
1
V = iR or i = v
R
Resistance is a dissipative property since v and i always have the same sign, and
the power is always positive
P = i
2
R = v
2
/R > 0
Energy always ows into a resistor, and cannot be recovered.
Summary: In modeling electrical systems we use three passive modeling elements:
The capacitor and inductor are energy storage elements in that power can ow into the
element, and recovered. The resistor is a dissipative element because the voltage an current
are algebraically related and power always ows into the element.
Sources: In addition to the three passive elements we will use two ideal electrical power
sources:
54
R
I (t) = I
+
-
s o
0
i
v
V
o
R increasi ng V i ncreasi ng
R = 0
o
I
V = RI
o o
o
o
I
The Voltage Source:
1
2
v (t)
v (t)
V (t)
1
2
+
-
s
v (t) - v (t) = V (t)
1 2
s
A voltage source will maintain the voltage at its terminals regardless of the current it
must supply to the connected load.
i
R
V (t) = V
+
-
s
o
0
i
v
V
o
R decreasi ng
i i ncreasi ng
R =
The Current Source:
1
2
v (t)
v (t)
I (t)
1
2
s
A current source will maintain the current supplied to the attached circuit regardless
of the voltage at it must generate to do so.
55
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
i
C
L
R
V (t)
+
-
s
i
i
1
2
3
i + -i + -i = 0
1 2 3

Spring Term 2008
Lecture 6
1
Reading:
Nise: Sec. 2.4 (pages 4555)
Sec. 1 (Introduction)
Sec. 4 (Electrical System Elements)
1 Modeling Electrical Systems (continued)
In Lecture 5 we examined the primitive electrical elements (capacitors inductors and resis
tors), and sources (voltage source and current source). We now look at how these elements
behave when connected together in a circuit.
Interconnection Laws:
(a) Kirchos Current Law (KCL): The sum of currents owing into(or out of) a
junction is zero. In the gure below, at the circled junction we sum the currents into
the junction to nd
i
1
i
2
i
3
= 0
We will dene a junction as a node, and if there are n circuit branches attached to a
node
n

i
n
= 0
i=1
where we dene the convention that positive current ow is into the node.
1
61
C
L
R
V (t)
+
-
s
Loop 1: v + v - V = 0
1
2
R C
s
Loop 2: v - v = 0
L C
Note: The + and - on the di agram
shows the di recti on of the
assumed vol tage drop.
+
+
+
-
-
-
Network
E
V
+
-
i
i
i
i
i
1
k
2
n-1
n
Kirchos Voltage Law (KVL): The sum of voltage drops around any closed loop in
a circuit is zero. The assumed sign convention for the voltage drop on each element
must be dened. Two clockwise loops are shown in the gure below. For loop (1)
v
R
+ v
C
V
s
(t),
while for loop (2)
v
L
v
C
(t) = 0.
Electrical Impedance:
Dene the impedance of an element or passive circuit as a transfer function relating current
I(s) to voltage V (s) at its terminals:
V (s)
Z(s) =
I(s)
In addition we can dene the admittance Y (s) as the reciprocal of the impedance:
1 I(s)
Y (s) = =
Z(s) V (s)
The Impedance of Passive Electrical Elements
62
+
-
C
i
8
i
+
-
L
8
i
-
+
R
8
branch
node Z
(a) The Capacitor:
For the capacitor
dv
i = C .
dt
Taking the Laplace Transform:
I(s) = CsV (s)
V (s) 1
Z
C
(s) = =
I(s) sC
or the admittance Y
C
(s) = sC.
(b) The Inductor:
For the inductor
di
v = L .
dt
Taking the Laplace Transform:
V (s) = LsI(s)
V (s)
Z
L
(s) = = sL
I(s)
or the admittance Y
L
(s) = 1/sL.
(c) The Resistor:
For the resistor
v = Ri.
Taking the Laplace transform
V (s) = RI(s)
V (s)
Z
R
(s) = = R
I(s)
or the admittance Y
R
= 1/R.
Impedance Nomenclature: We now introduce a graphical representation that will be
used to denote systems in many energy domains.
63
C
L
R
V (t)
+
-
s
Z
Z
Z
R
L
C
+
-
V (t)
s
Z
Z
1
2
+
-
V (t)
s
i
The impedance is drawn as a graph branch between two nodes. Nodes represent junctions
between the elements in the circuit. The arrow on the branch indicates both the assumed
direction of voltage drop across the element, and the assumed current direction.
Example 1
The electrical circuit, consisting of a capacitor C, an inductor L, and a resistor
R is shown below:
The impedance graph is shown on the right. The nodes on the graph represent
points of distinct voltage in the circuit.
Impedance Connection Rules
(a) Series connection: Two or more elements are dened to be connected in series if
they share a common current. For the two elements Z
1
and Z
2
in series below:
Using KCL at the junction between Z
1
and Z
2
:
i
Z
1
= i
Z
2
= i
Using KVL around the loop: v
Z
1
+ v
Z
2
V
s
= 0
V
s
= iZ
1
+ iZ
2
V (s)
Z
eq
= = Z
1
+ Z
2
I(s)
In general with n impedances Z
i
(i = 1, . . . , n) in series:
n
Z
eq
=

Z
i
i=1
64
C R L
Z
eq
Z Z
2
1
+
-
V (t)
s
i
=

Example 2
For the tree elements in series below:
1
Z
eq
= Z
C
+ Z
L
+ Z
R
= + sL + R
sC
or expressing the impedance as a transfer function (a ratio of polynomials):
V (s) LCs
2
+ RCs + 1
Z
eq
= =
I(s) Cs
(b) Parallel connection: Two or more elements are dened to be connected in parallel
if they share a common voltage. For the two elements Z
1
and Z
2
in parallel below:
Using KVL:
v
Z
1
= v
Z
2
= V
s
Using KCL at the node:
i
s
= i
Z
1
+ i
Z
2
1 I i
Z
1
+ i
Z
2
= =
Z
eq
V V
1 1 1
= +
Z
eq
Z
1
Z
2
In general for n impedances Z
i
(i = 1, . . . , n) in parallel, the equivalent impedance is:
n
1 1
Z
eq
Z
i
i=1
Alternatively, using admittances Y = 1/Z
n
y
eq
=
1
=

Y
i
.
Z
eq
i=1
Note: For N = 2 we can write
1 1 1 Z
1
+ Z
2
= + =
Z
eq
Z
1
Z
2
Z
1
Z
2
65
C
L
V (t)
+
-
s R
C
L, R
R
V (t)
+
-
s
Z = Ls
Z =
Z = R
L
4
+
-
V (t)
s
i ncl ude both i nductance and
resi stance of the coi l
L
Z = R
1
1
Cs
2
3
which leads to the very common representation
Z
1
Z
2
Z
eq
=
Z
1
+ Z
2
Example 3
Find the impedance of a capacitor C, and inductor L and a resistor R connected
in parallel:
1 1 1 1
= + +
Z 1/sC sL R
1 1
= sC + +
sL R
LCRs
2
+ Ls + R
=
RLs
V (s) RLs
Z = =
I(s) LCRs
2
+ Ls + R
Example 4
Find the impedance of the following circuit, assuming we should include resis
tance and inductance of the coil:
Z = Z
1
+ Z
4
(Z
2
+ Z
3
)
= Z
1
+
Z
4
(Z
2
+ Z
3
)
Z
4
+ Z
2
+ Z
3
= R +
(1/sC)(R
L
+ Ls)
1/sC + R
L
+ Ls
= R +
R
L
+ Ls
LCs
2
+ R
L
Cs + 1
66
v
Z
Z
1
2
+
-
V (t)
s
Z
1
v
Z
2
Z
Z
1
2
+
-
V (t)
s
C
V (t)
+
-
s
R
v
o
v
o
v = 0
V (s) RLCs
2
+ (RR
L
C + L)s + (R + R
L
)
Z = =
I(s) LCs
2
+ R
L
Cs + 1
The Voltage Divider: Consider two impedances in series with voltage V across them:
V (s)
I(s) =
(Z
1
+ Z
2
)
and
Z
2
V
Z
2
(s) = I(s)Z
2
= V
Z
2
(s) = V (s).
Z
1
+ Z
2
Similarly
Z
1
V
Z
1
(s) = V (s).
Z
1
+ Z
2
The voltage divider relationship may be used to nd the transfer function of many simple
systems.
Example 5
Find the transfer function relating V
0
to V
s
in the following circuit
Use the voltage divider relationship
Z
2
1/sC
V
0
= V
s
= V
s
Z
1
+ Z
2
R + 1/sC
V
0
(s) 1
H(s) = =
V (s) RCs + 1
Example 6
Find the transfer function relating V
0
to V
s
in the following circuit:
67
Z
Z
1
2
v
i
i
1
2
v = 0
I
I
(a)
C
L, R
R
V (t)
+
-
s
Z = Ls
Z =
Z = R
L
4
+
-
V (t)
s
as before, i ncl ude both
i nductance and resi stance
of the coi l
L
Z = R
1
1
Cs
2
3
v=0
v
V
s
o
v
o
Reduce the impedance graph to a series connection of two elements
Z
Z
1
5
+
-
V (t)
s
V
s
v
o
v=0
Z
5
V
0
= V
s
Z
1
+ Z
5
where
Z
4
(Z
2
+ Z
3
)
Z
5
= Z
4
(Z
2
+ Z
3
) =
Z
4
+ Z
2
+ Z
3
(1/sC)(R
L
+ L
s
)
=
1/sC + R
L
+ Ls
Using the voltage divider relationship, the transfer function is
R
L
+Ls
V
0
(s) Z
5 LCs
2
+R
L
Cs+1
H(s) = = =
R
L
+L
s
V
s
(s) Z
1
+ Z
5 R
1
+
LCs
2
+R
L
Cs+1
R
L
+ Ls
H(s) =
R
1
LCs
2
+ (R
1
R
L
C + L)s + (R
1
+ R
L
)
The Current Divider: Consider two impedances in parallel:
68
C L
R
It
=
I
i
1
i
2
v t
o
>
Z
Z
1
2
i
i
1
2
v = 0
I
I(s)
Z
3
(a)
(b)
V (s)
o
V (s)
a
Using KCL at the top node (a),
I i
1
i
2
= 0 or i
1
+ i
1
= I
But i
1
= V/Z
1
, and i
2
= V/Z
2
so that
V
Z
1
+
V
Z
2
= I or V =
1
1/Z
1
+ 1/Z
2
I
V 1/Z
1
Y
1
i
1
= = I = I
Z
1
(1/Z
1
+ 1/Z
2
) Y
1
+ Y
2
Similarly
Y
2
i
2
= I.
Y
1
+ Y
2
The current divider may be used to nd transfer functions for some simple circuits.
Example 7
Find the transfer function
V
o
(s)
H(s) =
I(s)
in the following circuit:
Draw the system as an impedance graph:
69
Let Z
1
= 1/sC, Z
2
= R, and Z
3
= sL. We will use V
o
(s) = I
2
(s)Z
3
(at node
(b)), and nd I
2
(s) from the current division at node (a):
1
1
I
2
(s) =
1
Z
2
+Z
3
1
I(s) = I(s)
+ (1/Z
1
)(Z
2
+ Z
3
) + 1
Z
1
Z
2
+Z
3
1 1
= I(s) = I(s)
Cs(R + Ls) + 1 LCs
2
+ RCs + 1
Ls
V
o
(s) = I
2
(s)Ls = I(s)
LCs
2
+ RCs + 1
or
V
o
(s) Ls
H(s) = =
I(s) LCs
2
+ RCs + 1
610
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
C
V (t)
+
-
s
R
R
C
(a)
(b)
1
1
2
2
v
o
H(s) =
V (s)
V (s)
o
s
(c)
Z
Z
1
2
+
-
s
(a) (b)
V (s)
(c)

Spring Term 2008
Lecture 7
1
Reading:
Nise: Sec. 2.4
1 Transfer Function Generation by Simplication
Example 1
Find the transfer function for a lead-lag compensator
Draw as
where
1 R
1
Z
1
=
C
1
s
R
1
=
R
1
C
1
s + 1
1 R
2
Z
2
=
C
2
s
R
2
=
R
2
C
2
s + 1
1
71
C
L
I (s)
1
R
2
R
V (s)
L
i
i
1
2
=
Z
Z
1
2
(a)
i
i
1
2
I
I
Using the voltage divider formed by Z
1
and Z
2
Z
2
V
o
= V
s
Z
1
+ Z
2
(
R
2
C
R
2
2
s+1
)
= V
s
(
R
1
C
R
1
1
s+1
) + (
R
2
C
R
2
2
s+1
)
R
2
R
1
C
1
s + R
2
= V
s
R
1
R
2
(C
1
+ C
2
)s + (R
1
+ R
2
)
V
o
(s) R
2
R
1
C
1
s + R
2
H(s) = =
V
s
(s) R
1
R
2
(C
1
+ C
2
)s + (R
1
+ R
2
)
Example 2
Find the transfer function
V
L
(s)
H(s) =
I(s)
for the circuit:
We note that
V
L
(s) = Z
L
i
2
(s) = Lsi
2
(s)
Combine elements to let
1 1
Z
1
= = R
1
+
Y
1
Cs
1
Z
2
= = R
2
+ Ls
Y
2
and use the current divider relationship at node (a).
72
C
1
R
2
R
L L
1 2
+
-
V
Y
2
i
2
(s) =
Y
1
+ Y
2
where
1 sC
Y
1
= =
Z
1
sCR
1
+ 1
1 1
Y
2
= = .
Z
2
sL + R
2
Then
R
1
Cs + 1
i
2
(s) = I(s)
Cs(R
2
+ Ls) + (R
1
Cs + 1)
and
LCR
1
s
2
+ Ls
V
L
(s) = Lsi
2
(s) = I(s)
LCs
2
+ C(R
1
+ R
2
)s + 1
so that the transfer function is
V
L
(s) LCR
1
s
2
+ Ls
H(s) = =
I(s) LCs
2
+ C(R
1
+ R
2
)s + 1
2 Transfer Function Generation through Mesh (Loop) Currents
This method expresses the system dynamics as a set of simultaneous algebraic equations in a
set of internal mesh (or loop) currents. It is useful for complex circuits containing a voltage
source.
The following example sets out the method in a series of steps.
Example 3
Find the transfer function of a bridged-T lter:
Note: This is a dicult example to solve using impedance reduction methods.
It is, however, well suited to the mesh current method.
Draw the system as an impedance graph:
73
Z
Z Z
Z
Z
2 1
5
3
4
V
1
3
2
Dene a set of (clockwise) loops as shown, ensuring that every graph branch is
covered by at least one loop. The loops 1, 2 and 3 are somewhat arbitrary. We
assume hypothetical continuous mesh currents i
1
, i
2
, and i
3
that ow around
each loop.
Step 1: Write loop (compatibility) equations for each loop (using the arrows
on the graphs branches to dene the direction of the voltage drop):
V
Z
1
+ V
Z
3
V = 0
V
Z
2
+ V
Z
4
V
Z
3
= 0
V
Z
5
+ V
Z
2
V
Z
1
= 0
Step 2: Dene the mesh currents i
1
, i
2
, and i
3
and write the current in each
branch in terms of the mesh currents (use the arrows on the loops to dene
the signs):
i
Z
1
= i
1
i
3
i
Z
2
= i
2
i
3
i
Z
3
= i
1
i
2
i
Z
4
= i
2
i
Z
5
= i
3
Step 3: Write the mesh equations in terms of the mesh currents
Z
1
(i
1
i
3
) + Z
3
(i
1
i
2
) = V
Z
2
(i
2
i
3
) + Z
4
i
2
Z
3
(i
1
i
2
) = 0
Z
5
i
3
Z
2
(i
2
i
3
) Z
1
(i
1
i
3
) = 0
which is a set of 3 simultaneous algebraic equations in the loop currents i
1
,
i
2
, and i
3
.
Step 4: We note that the output is V
Z
4
= i
2
Z
4
, therefore solve for i
2
and
create the transfer function in terms of the impedances.
74
V
+
-
R
R
C
1
2
v
o
V
Z
1
Z
2
Z
3
1
2
Example 4
Use the mesh current method to nd the transfer function
V
o
(s)
H(s) =
V (s)
in the lead network:
Let Z
1
= 1/Cs, Z
2
= R
1
, and Z
3
= R
2
.
Step1: The loop equations are:
V
Z
2
+ V
Z
3
V = 0
V
Z
1
V
Z
2
= 0
Step 2: The mesh currents are i
1
and i
2
, and
i
Z
1
= i
2
i
Z
2
= i
1
i
2
i
Z
3
= i
1
Step 3: Rewrite the loop equations
Z
2
(i
1
i
2
) + Z
3
i
1
= V
Z
1
i
2
Z
2
(i
1
i
2
) = 0
or
(Z
2
+ Z
3
)i
1
Z
2
i
2
= V
Z
2
i
1
+ (Z
1
+ Z
2
)i
2
= 0
In matrix form

Z
2
+ Z
3
Z
2

i
1

V

=
Z
2
Z
1
+ Z
2
i
2
75
0
L
R
2
C
1
C
L
1 1 0v ac
power suppl y/recti fi er fi l ter
l oad
2
3
1
J J
J
L J
!
L J
L J
At :
At :
! At :
ri ppl e
Step 4: V
o
= i
1
Z
3
, so we solve for i
1
(using any method). Using Cramers Rule:
V Z
2
0 Z
1
+ Z
2 Z
1
+ Z
2
i
1
= = V
Z
2
+ Z
3
Z
2
Z
1
Z
2
+ Z
1
Z
3
+ Z
2
Z
3

Z
2
Z
1
+ Z
2

so that
(Z
1
+ Z
2
)Z
3
V
o
= V
Z
1
Z
2
+ Z
1
Z
3
+ Z
2
Z
3
and
V
o
(s) (R
1
+ 1/Cs)R
2
H(s) = =
V (s) R
1
/Cs + R
2
/Cs + R
1
R
2
R
1
R
2
Cs + R
2
H(s) =
R
1
R
2
Cs + (R
1
+ R
2
)
Note that this problem could have be done using a voltage-divider
approach
Example 5
A common full-wave wave rectied dc power supply for electronic equipment is:
Typical waveforms in the circuit are
76
L
R
2
C
C
L
Theveni n source model F-secti on fi l ter
l oad
+
-
V
R
1
'
(a) (b) (c)
(d)
v
o
1
Z Z
Z
Z
2 1
4
3
V
=
> ?
@
Z Z
Z
Z
2 1
4
3
V
1
2
The lter acts to smooth the full-wave rectied waveform at (2) to produce a
dc output at (3) with a very much reduced ripple.
We use a linearized model of the transformer/rectier circuit, with a voltage
source (with a waveform as at (2) above) and a series resistor R (a Thevenin
source) - see Lecture 8 - as below:
The task is to nd the transfer function
V
o
(s)
H(s) = .
V (s)
Solution: Combine series and parallel impedances to simplify the structure.
Draw as an impedance graph
where
Z
1
= R
1
Z
2
= sL
1 R
L
/sC
2
Z
3
=
sC
2
R
L
=
R
L
+ 1/sC
2
1
Z
4
=
sC
1
The system output is V
Z
3
= i
Z
3
Z
3
.
Choose mesh loops to contact all branches as below.
77
The loop equations are:
v
Z
1
+ v
Z
4
V = 0
v
Z
2
+ v
Z
2
v
Z
4
= 0
and the branch currents (in terms of the mesh currents) are:
i
Z
1
= i
1
i
Z
2
= i
2
i
Z
3
= i
2
i
Z
4
= i
1
i
2
Rewrite the loop equations in terms of the mesh currents:
Z
1
i
1
+ Z
4
(i
1
i
2
) = V
Z
2
i
2
+ Z
3
i
2
Z
4
(i
1
i
2
) = 0
or
(Z
1
+ Z
4
)i
1
Z
4
i
2
= V
Z
4
i
1
+ (Z
2
+ Z
3
+ Z
4
)i
2
= 0
giving a pair of simultaneous algebraic equations in the mesh currents.
Solve for i
Z
3
= i
2

Z
1
+ Z
4
Z
4

i
1

=

V

Z
4
Z
2
+ Z
3
+ Z
4
i
2
0
and using Cramers Rule:
(Z
1
+ Z
4
) V
Z
4
0
Z
4
V
i
2
= =
(Z
1
+ Z
4
)
(Z
1
+ Z
4
)(Z
2
+ Z
3
+ Z
4
) Z
2
Z
4 4
Z
4
Z
2
+ Z
3
+ Z
4
and since V
o
= Z
3
i
2
V
o
Z
3
Z
4
H(s) = =
V (s) Z
1
Z
2
+ Z
1
Z
3
+ Z
1
Z
4
+ Z
2
Z
4
+ Z
3
Z
4
Substituting the impedances of the branches
R
L
H(s) =
LC
1
C
2
R
o
R
1
s
3
+ L(C
1
R
1
+ R
o
C
2
)s
2
+ R
o
R
1
(C
1
+ C
2
)s + (R
1
+ R
2
)
78
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
L
E
L
E
8
1

a vol tage source
a current source
vol tage i s i ndependent of
the current suppl i ed to the
l oad.
current i s i ndependent of
the vol tage requi red to
mai ntai n the current.
+
-
8J system 8
+
1J
1
system
+
-
battery
R
L
i
v
R
L
v
v

Spring Term 2008
Lecture 8
1
1 Electrical System Modeling (continued)
Modeling Real Sources: Up to this point we have considered only ideal voltage and
current sources:
However, practical (real) sources do not have an ideal characteristic they are power limited.
For example we may think of a battery as an ideal voltage source - say with a 9 volt output,
but if we were to measure the terminal voltage under increasing load currents
we would nd that at high currents the voltage decreases in a nonlinear manner:
1
81
+
-
system
v
i
R
o
s
source model
v = V - iR
o s
v
i
V
0
i
R = =
dv
di
actual source
characteri sti c
o
v
i
oc
sc
oc
sc
Theveni n source model
L
E
8
"l i near" regi on

open-ci rcui t vol tage
when E = (4 = ).
short-ci rcui t current
when L = (4 = ).
For a real source, the vol tage
decreases as the current
i ncreases
The voltagecurrent relationship is determined by the electrochemical reaction within the
battery. It may be possible to dene a region of operation at low currents in which there is
an approximately linear relationship between voltage and current
v(i) v
oc
R
s
i
where v
oc
is the open-circuit voltage (when i = 0), and R
o
is a resistance.
How can we model this linear part of the characteristic?
One way to do this is to model the real source(the battery) as an ideal voltage source V
s
in series with a resistor R
o
(the value of which is found experimentally). The voltage drop
v
R
o
= iR
o
across the resistor accounts for the droop in the source characteristic.
This is known as the Thevenin equivalent source model.
82
i ndi cated vol tage
i
R
L
v
out
1 . 234
1 . 234
R
L
actual
vol tage
ampl i fi er
50 9
v
out
v
1
vol tmeter
Functi on Generator
v =
out
R
R + 50
L
L
v
1
The meter i s cal i brated to read
correctl y only when a 50 9 l oad
resi stor i s connected!
Note that
(1) The open-circuit (i = 0) voltage is V
s
.
(2) The short-circuit (v = 0) current is i
sc
= V
s
/R.
Example 1
Electronic instruments are often designed to have a specic output impedance
(resistance) R
o
. For example when the Tektronix function generator in the 2.004
laboratory is connected to an arbitrary load resistance R
L
the measured output
voltage v
out
will not necessarily be equal to the value set on the front panel:
(a) If R
L
= (no load connected) the true voltage at the terminals is twice the
indicated value.
(b) If R
L
= 0 (short circuit) the actual voltage is zero but the indicated voltage
is unchanged.
(c) If R
L
= 50 the indicated and measured output voltages are the same.
Why?
Answer: The function generator is specied as having an output impedance of
50. In fact, if you look at the circuit schematic, you will nd that there is a
50 resistor in series with the output terminal:
83
system
v R
o
N
Norton source model
(a)
N
l
R
o
In the lab we have 50 terminator resistors connected across the output terminals
so that the voltage reads correctly.
A second approach to modeling the real source characteristic
v
i
V
0
i
R = =
dv
di
o
L
E
oc
sc
oc
sc
v = V - i R
oc L
L
o
is to use a current source I
N
in parallel with the resistor R
o
This is known as the Norton equivalent source model.
If no load resistor is connected (open-circuit)
v
os
= I
N
R
o
.
The short-circuit current is
i
sc
= I
N
.
If the load is a resistor R
L
R
L
R
o
R
L
v/i
v = = v
oc
=
R
L
+ R
o
R
L
+ R
o
v + R
o
and rearranging
v = v
oc
iR
L
.
The Norton and Thevenin source models are equivalent and cannot be dis
tinguished by measurements at the terminals.
84
R = 5 9
o
V = 200 v
th
+
-
Example 2
Find Thev
Thevenin Model:
Measurements on an industrial dc power supply showed it to have an open-circuit
voltage of 200 v, but when 10 resistor is connected, the voltage drops to 133v.
enin and Norton equivalent source models.
R
o
R
L
V
th
+
-
I
N
R
o
R
L
R = 10 9
L
R
o
V = 200 v
th
+
-
V
th
= v
oc
= 200 v.
Using the voltage divider relationship
10
v
o
= v
R
L
= = 133volts
10 + R
o
which gives
R
o
= 5.
The Thevenin model is:
Norton Model: From above, the short-circuit current is
V
th
200
i
sc
= = = 40 amps
R
o
5
and the Norton model is:
85
R = 5 9
N
I = 40 A
N
86
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
system
Z
L
Z
Z
o
L
Z
Z
o
L
I
N
Theveni n
Norton
th
V
Spring Term 2008
Lecture 9
1
Reading:
Nise: Sec. 2.5
1 Thevenin and Norton Source Models
We state the following without proof:
Any linear system (regardless of its internal complexity) containing a single source (voltage
or current), and with an external load element Z
L
, may be modeled as either
(a) A voltage source V
th
with a series impedance Z
o
(a Thevenin equivalent circuit), or
(b) A current source I
N
with a parallel impedance Z
o
(a Norton equivalent circuit).
The arbitrary system shown above can be represented in either form, where
Z
o
is the system output impedance, and is the same in each case.
V
th
1
copyright
is the Thevenin source voltage
circuit terminal voltage.
c D.Rowell 2008
found by removing Z
o
and measuring the open
91
system
short-ci rcui t
current
N
I
C
V (t)
+
-
s
R
Z
L
Z =
o
C
+
-
R
R
RCs + 1`
Z
L
R
RCs + 1
1
RCs + 1
V
s
JD
8 system
open-ci rcui t
vol tage
I
N
is the Norton source current found by short-circuiting the output and measuring
the current.
To nd the output impedance Z
o
set the internal source to zero, and measure the input
impedance at the systems output terminals. Some care must be taken in setting any source
to zero:
(a) To set a voltage source to zero, short-circuit it, for example to nd Z
o
in the circuit
below:
1 R
Z
o
=
Cs
R =
CRs + 1
The Thevenin source voltage is found by recognizing the left-hand circuit as a voltage-
divider and nding the terminal voltage:
1/(Cs) 1
V
th
= V
s
= V
s
R + 1/(Cs) RCs + 1
The complete Thevenin equivalent for this system is:
92
C s
R
Z
L I
C s
R
I
Z =
o
1
Cs
Z
L
I
N
1
Cs
C
V (t)
+
-
s
R
C
V (t)
+
-
s
R
L
V
o
(b) To set a current source to zero, remove it from the circuit as shown below:
In this case the output impedance seen at the output terminals is
1
Z
o
=
Cs
and the Norton short circuit current is simply
I
N
= I
(Note that R does not appear in the system formulation.) The Norton equivalent is:
Example 1
Use Thevenin and Norton source models to nd the transfer function of the
following system when the load is an inductor L.
V
o
(s)
H(s) =
V (s)
Thevenin Model: For the source model on the left
1/(Cs) 1
V
th
= =
R + 1/(Cs) RCs + 1
1 R
Z
o
=
Cs
R =
CRs + 1
Then for the full system on the right
93
C
R
i =
sc
V
R
+
-
V C
R
Z =
o
R
RCs + 1
+
-
V
Ls I =
N
V
R
R
RCs + 1
Ls
R
RCs + 1
1
RCs + 1
V
V
o
0
s
Using a voltage divider relationship
Ls
V
o
(s) = V
th
(s)
Ls + R/(CRs + 1)
Ls(RCs + 1) 1
= . V (s).
RLCs
2
+ Ls + R RCs + 1
The transfer function is
V
o
(s) Ls
H(s) = =
V (s) RLCs
2
+ Ls + R
Norton Model: For the Norton model
1 1 R
I
N
= i
sc
= V, Z
o
= R =
R Cs RCs + 1
and the equivalent system model is
Then the transfer function is found from
V
o
(s) =
Z
o

1
Cs

I
N
(s)
RLs 1
= V (s)
Ls + R/(RCs + 1) R
Ls
= V
RLCs
2
+ Ls + R
94
F m
v
Vel oci ti es and forces are
measured wi th respect to
an i nerti al reference frame
As with the Thevenin method, the transfer function is
V
o
(s) Ls
H(s) = =
V (s) RLCs
2
+ Ls + R
2 Modeling Mechanical Systems
We now move our attention to deriving models of mechanical systems with motion in one
dimension. The approach will be to draw analogies with the electrical modeling methods so
as to have a unied technique that can be applied without regard to the energy domain.
(1) Choice of Modeling Variables: As in the electrical domain, we select a pair of
variables whose product is power, that is we choose F force (N), and v velocity
(m/s) since
P = Fv
(2) Modeling Elements: As in the electrical domain we nd there are two energy storage
elements, and one dissipative element.
(a) The mass element.
For a mass element m
dp dv
m
F
m
= = m
dt dt
where p is the momentum. We will use the elemental relationship
dv 1 1
= F or V (s) = F (s)
dt m ms
The energy stored in a moving mass element is
t

1
2
E = Pv dt = mv
2

which is a function of velocity v and is stored as kinetic energy.
Note: Inertial forces and velocities must be measured with respect to a non-
accelerating inertial reference frame. In this course we will assume a reference
velocity v
ref
= 0.
95
. .
L
(b) The spring (compliance) element:

K
F
F
x
x
v
v 1
2
2
1
Let x
spring
= (x
2
x
1
) x
o
where x
o
is the rest length of the spring. Then
dx
spring
dx
2
dx
1
v
spring
=
dt
=
dt

dt
= v
2
v
1
From Hookes law,
F = Kx
spring
where K is the stiness (N/m).
dF
K
K
= Kv
K
or F
K
(s) = v
K
(s)
dt s
The energy stored in a moving mass element is
t

1
E = Pv dt = F
K
2
2K

which is a function of velocity F
K
and is stored as potential energy.
(c) The viscous friction (dissipative) element: (Also known as the dashpot or
damper element)
Let V
B
= v
2
v
1
, the elemental equation is
F
B
= Bv
B
or F
B
(s) = Bv
B
(s)
The power ow is
P = Fv = Bv
2
0
so that the power ow is uni-directional and cannot be recovered.
(3) Ideal Sources: We dene a pair of ideal sources
(a) Force Source: The force source maintains a prescribed force F
s
regardless of
the velocity at which it travels.
96
v
V
0
a velocity source
vel oci ty i s i ndependent of
the force appl i ed to the
l oad.
F
s
V
s
Arrow i s i n the di recti on
of the assumed vel oci ty
drop
m F
F
F
3
2
1
v
F + F - F = m
1 2 3
dv
dt
v
F
F
0
a force source
force i s i ndependent of
the vel oci ty requi red to
mai ntai n the current.
s
F
s
Force acts to accel erate
the node at the head of
the arrow i n the posi ti ve
vel oci ty di recti on
(b) Velocity Source: The velocity source maintains a prescribed velocity V
s
re
gardless of the force required to maintain that velocity.
Interconnection Rules:
(a) Continuity: Consider a mass element with n external forces acting on it:
n

F
i
= m
dv
m
dt
i=1
or
n

F
i
m
dv
m
= 0
dt
i=1
97
m F
F
F
3
2
1
v
F
3
F
m
F
1
F
2
(+ve) (+ve)
(-ve) (-ve)
node represents the moti on
of the mass el ement
m m
K
B
v v
1
1 2
2
m m
K
B
v v
1
1 2
2
F F
F
F
K K
B B
v
v
1 2
K
B
m
m
1
2
v = 0
F + F - F = 0
B K m
2
-F - F - F = 0
B K m
1
Note: On a branch representi ng
a mass el ement, the arrow always
poi nts to the i nerti al reference node.
If we substitute a ctitious force F
m
= m dv
m
/ dt (known as a dAlembert force) we
van write
n

F
i
F
m
= 0
i=1
and state the following
The sum of all forces acting on a mass element (including the dAlembert
force) is zero.
If we represent a point in space containing a mass element as a node on a graph, and
the forces acting on the node as branches, where an arrow pointing at the node means
that positive force acts to accelerate the node in the reference direction.
We can write
F
1
+ F
2
F
3
= F
m
= 0
The node represents a point in a mechanical system with a distinct velocity.
The continuity condition for a mechanical system is equivalent to Kirchos
current law in an electrical system.
Example 2
The system
may be drawn as a graph:
98
K
B
i deal spri ngs and dampers are
assumed to be massl ess
K B
v
v
F - F = 0
K B
Notice the arrow directions on the branches - as implied by the right hand
gure above. The graph implies a pair of continuity equations at the two
nodes:
dv
1
F
K
F
B
F
m
1
= 0 or F
K
F
B
= m
1
dt
dv
2
F
K
+ F
B
F
m
2
= 0 or F
K
+ F
B
= m
2
dt
For a massless node (that is at the interconnection of n ideal massless elements)
n

F
i
= 0
i=1
For example
implies F
K
F
B
= 0 at the junction.
99
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
m m
K
1 2
m
3
= > ?
K 1
2
v
a v
c
v
b
v = 0
ref
!

Spring Term 2008
Lecture 10
1
Reading:
Nise: Sec. 2.5 (pages 5966)
1 Modeling Mechanical Systems (continued)
In the previous lecture we examined the node (continuity) equations for mechanical systems.
Compatibility Condition: Consider the following system:
We dene the velocity drop between two points in a system as the dierence in the velocities
(measured with respect to the reference velocity), at two points in the system. For example
v
ab
= v
a
v
b
.
If we move from node to node around a loop, ending at the starting node, summing the
velocity drops as we go, for example a loop from (a) (b) (c) (a), and sum the velocity
drops
v
ab
+ v
bc
+ v
ca
= (v
a
v
b
) + (v
b
v
c
) + (v
c
v
a
) = 0
The graph for the above system is
1
101
m m
K
B
v v
1
1 2
2
B B
1 2
3
F
v
1
v
2
K
m
1
B
1
B
2
m
2
F
B
3
(a)
(b)
(c)
1
2
3
4
v = 0
and around the three loops
v
K
1
+ v
m
2
= (v
a
v
b
) + (v
b
v
d
) (v
a
v
d
) = 0 (Loop 1) v
m
3
v
K
2
+ v
m
3
= (v
b
v
c
) + (v
c
v
d
) (v
b
v
d
) = 0 (Loop 2) v
m
2
v
K
1
+ v
K
2
+ v
m
3
= (v
a
v
b
) + (v
b
v
c
) + (v
c
v
d
) (v
c
v
d
) = 0 (Loop 3) v
m
3
In all cases the sum of the velocity drops is zero. The compatibility condition for mechanical
systems states:
The sum of velocity drops, from node to node around any closed loop on a
system graph is zero.
which is analogous to Kirchos voltage law for an electrical system.
Example 1
Write compatibility equations, and continuity equations for the following system:
The system graph, with four loops dened, is
The compatibility equations for the four loops are:
v
m
2
v
B
2
= 0 (Loop 1)
v
B
2
v
B
1
+ v
B
3
= 0 (Loop 2)
v
m
1
v
B
2
= 0 (Loop 3)
v
K
v
B
3
= 0 (Loop 4)
Continuity equations at nodes (a) and (b) are
F (t) F
K
F
B
3
F
B
1
F
m
1
= 0 (Node(a))
F
K
+ F
B
3
F
B
1
F
m
2
= 0 (Node(a))
102
Notes:
1) Branches associated with mass elements always connect to the inertial reference node
because (i) forces are measured with respect to the inertial reference frame, and (ii)
velocities are measured with respect to the inertial reference frame.
2) The arrow on a branch associated with a mass element always points away from the
node (toward the reference node) because of the sign on the dAlembert force in the
continuity equation
n

F
i
F
m
= 0.
i=1
Analogy with Electrical Systems: We can compare the interconnection rules dened by
system graphs for electrical and mechanical systems:
Electrical: Currents into a junction (node) sum to zero (KCL).
Mechanical: Forces at a point (node) (including the dAlembert force) sum to zero.
Electrical: Voltage drops around a closed loop sum to zero (KVL).
Mechanical: Velocity drops around a closed loop in a mechanical system sum to
zero.
We use these similarities to make the following analogies between variables in the two energy
domains:

voltage velocity
electrical

mechanical
current force
Note: The opposite analogies can be made, and are in fact used by many authors, however
the above grouping is particularly convenient for use in the graph based method we are
developing.
Mechanical Impedance: With the above analogy we dene the mechanical impedance as
the ratio of velocity to force
V (s)
V (s)
Z
mech
= compare with Z
elect
= .
F (s) Is
We can also dene mechanical admittance
1 F (s)
Y
mech
= =
Z
mech
V (s)
Elemental Impedances:
(a) mass element:
V
J
(s) 1
F
J
(s) = msV
J
(s) Z
J
= =
F
J
(s) ms
103
m
K
B
v
m
F
v = 0
1
ms
s
K
1
B
v
m
F
v = 0
(b) spring element:
V
K
(s) s
sF
K
(s) = KV
K
(s) Z
K
= =
F
K
(s) K
(c) damper element:
V
B
(s) 1
F
B
(s) = BV
B
(s) Z
B
= =
F
B
(s) B
The rules for combining series and parallel mechanical impedances are the same as for elec
trical impedances, leading to the same methods for generating transfer functions.
Example 2
Use impedance methods to nd the transfer function
V (s)
H(s) =
F (s)
for
The impedance of the three passive elements is
1 1 1 1 K ms
2
+ Bs + K
= + + = ms + + B =
Z
eq
Z
J
Z
K
Z
B
s s
V (s) s
H(s) = = Z
eq
=
F (s) ms
2
+ Bs + K
Example 3
Find the transfer function relating the velocity of the mass m
2
to the input force
F (t) in the system:
104
m m
K
B
v v
1
1 2
2
B B
1 2
3
F
The system graph on the left may be simplied:
v
1
v
2
s
K
1
m s
1
1
B
1
1
B
2
2
F
1
B
3
= >
?
v = 0
1
m s
v
1
v
2
F
= >
?
v = 0
Z
Z
Z
2
1
3
where
Z
1
=
s
K

1
B
3
Z
2
=
1
sm
2

1
B
2
=
=
s
B
3
s + K
1
m
2
s + B
2
1 1 1
Z
3
=
sm
1
B
1
=
m
1
s + B
1
We need to compute the velocity at node (a)
Z
3
(Z
1
+ Z
2
)
V
1
(s) = F (s)Z
eq
= F (s) (Z
3
(Z
1
+ Z
2
)) = F (s)
Z
1
+ Z
2
+ Z
3
The output velocity V
2
(s) can be found using the velocity divider
Z
2
Z
2
Z
3
V
2
(s) = V
1
(s) = F (s)
Z
1
+ Z
2
Z
1
+ Z
2
+ Z
3
V
2
(s) Z
2
Z
3
B
3
s + K
H(s) = = =
F (s) Z
1
+ Z
2
+ Z
3
a
3
s
3
+ a
2
s
2
+ a
1
s + a
0
where symbolic software was used to nd
a
3
= m
1
m
2
a
2
= m
1
(B
2
+ B
3
) + m
2
(B
1
+ B
3
)
a
1
= K(m
1
+ m
2
) + B
1
B
2
+ B
1
B
3
+ B
2
B
3
a
0
= K(B
1
+ B
2
)
105
v
1
v
2
s
K
1
m s
1
1
B
1
1
B
2
2
F
1
B
3
= >
?
v = 0
1
m s
v
1
v
2
F
= >
?
v = 0
Z
Z
Z
2
1
3
Example 4
Repeat Example 4 using node equations to nd H(s).
Solution: From Example 4, the impedance graph, and a simplied form are
where
Z
1
=
s
K

1
B
3
Z
2
=
1
sm
2

1
B
2
=
=
s
B
3
s + K
1
m
2
s + B
2
1 1 1
Z
3
=
sm
1
B
1
=
m
1
s + B
1
Write a pair of node equations expressing the continuity conditions at (a) and
(b):
F F
Z
1
F
Z
2
= 0 at node (a)
F
Z
2
F
Z
3
= 0 at node (b)
For convenience, use admittances instead of impedances. Let Y
1
= 1/Z
1
, Y
2
=
1/Z
2
, and Y
3
= 1/Z
3
. Substitute for the admittance relationships (F = Y v) on
each branch:
Y
1
(v
a
0) + Y
2
(v
a
v
b
) = F at node (a)
Y
2
(v
a
v
b
) Y
3
(V
b
0) = 0 at node (b),
which are a pair of simultaneous linear equations in v
a
and v
b
:

Y
1
+ Y
2
Y
2

v
a

F

=
Y
2
(Y
2
+ Y
3
) v
b
0
106
and may be solved using Cramers rule:
Y
1
+ Y
2
F
Y
2
0 Y
2
F (s)
v
b
(s) = v
m
1
(s) = =
Y
1
+ Y
2
Y
2
Y
1
Y
2
+ Y
2
Y
3
+ Y
3
Y
1
Y
2
(Y
2
+ Y
3
)
Z
1
Z
3
= F (s)
Z
1
+ Z
2
+ Z
3
by dividing numerator and denominator by Y
1
Y
2
Y
3
. Then
x
m
1
(s) 1 v
m
1
(s) 1 Z
1
Z
3
H(s) = = = ,
F (s) s F (s) s Z
1
+ Z
2
+ Z
3
which is the same result found using ad-hoc impedance reduction methods in
Example 4.
107
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
F
m
m
m
B
B
B
K
1 1
2
2
3
3
v
m
1
F
m
m m
B
B
B
K
1
1
2
2
3
3
=
>
?
m
v
1

Spring Term 2008
Lecture 11
1
Classroom Example: Symbolic Transfer Function Generation
Using MATLAB and Maple
Consider the mechanical system shown below:
The task is to nd the transfer function relating the velocity of mass m
1
to the input force
F (s) using the symbolic math package Maple.
The system linear graph is shown below:
We start with the node equation method:
(a) Write the continuity equations at the nodes (a), (b), and (c):
F F
B
1
F
B
2
F
B
3
F
m
3
= 0
F
B
2
F
K
F
m
2
= 0
F
B
1
+ F
K
F
m
1
= 0
1
111
(b) Substitute for the elemental forces using elemental impedance relationships (in this case
it is more convenient to use admittances Y = 1/Z) and velocities at the nodes.
F Y
B
1
(v
a
v
c
) Y
B
2
(v
a
v
b
) Y
B
3
v
a
Y
m
3
v
a
= 0
Y
B
2
(v
a
v
b
) Y
K
(v
b
v
c
) Y
m
2
v
b
= 0
Y
B
1
(v
a
v
c
) + Y
K
(v
b
v
c
) Y
m
2
v
c
= 0
where Y
m
1
= m
1
s, Y
m
2
= m
2
s, Y
m
3
= m
3
s, Y
B
1
= B
1
, Y
B
2
= B
2
, Y
B
3
= B
3
, and
Y
K
= K/s.
(c) These three linear algebraic equations are used to solve for v
a
= v
m
1
, which denes the
transfer function.
While the solution is easy in principle, the algebraic manipulations become messy and so we
look at the use of symbolic math software contained in MATLAB and Maple to assist in the
solution:
Using MATLABs Symbolic Toolbox,
Using Maple directly, and
Using the software package Maple Syrep.
Using MATLABs Symbolic Toolbox to Find the Transfer Function:
The Symbolic Toolbox is an optional add-on to MATLAB. It is a sub-set of Maple, and
allows the symbolic solution of linear algebraic equations, such as required in this example.
The following is a m le, Lecture11Example.m that illustrates how MATLAB can be used:
% 2.004 Spring 2008
% Classroom Example: Symbolic Transfer Function Generation
% Using MATLABs Symbolic Toolbox.
%
%---------------------------------
% Declare all of the symbolic variables as syms. The specifier real
% at the end says that we expect all of these variables to be real.
syms s F va vb vc YB1 YB2 YB3 Ym1 Ym2 Ym3 YK B1 B2 B3 m1 m2 m3 K real
% Solve for the three nodal velocities using the nodal equations
[va, vb, vc] = solve(F-YB1*(va -vc) - YB2*(va -vb) - YB3*va - Ym3*va = 0,...
YB2*(va-vb) - YK*(vb -vc) - Ym2*vb = 0,...
YB1*(va-vc) + YK*(vb-vc) - Ym1*vc = 0);
% The output variable of interest is vc. Substitute for the impedance of
% all of the elements (Note - we could have made the substitutions directly
% into the equations above and saved these steps):
vc=subs(vc,YB1,B1);
vc=subs(vc,YB2,B2);
vc=subs(vc,YB3,B3);
112
vc=subs(vc,Ym1,m1*s);
vc=subs(vc,YK,K/s);
% THe transfer function is vc/F
H=vc/F;
% The rest is "fluff". Use "simplify(H)" to make H a rational fraction,
% use "collect(H,s)" to collect all of the coefficients for each power of s,
% use "sort(H)" to order the terms in descending powers of s.
H = sort(collect(simplify(H),s))
The following is a MATLAB command line fragment showing the output from the .m le,
and how numerical values are substituted into the result:
>>
>> Lecture11Example
H =
(s^2*B1*m2+s*B1*B2+B1*K+B2*K)/(s^4*m1*m3*m2
+(B3*m1*m2+m1*B1*m2+m1*B2*m3+m1*B2*m2+B1*m3*m2)*s^3
+(B3*m1*B2+B3*B1*m2+m1*B1*B2+m1*m3*K+B1*B2*m3+B1*B2*m2+m3*K*m2)*s^2
+(B3*m1*K+B3*B1*B2+B3*K*m2+m1*B1*K+m1*B2*K+B1*m3*K+B1*K*m2+B2*m3*K+B2*K*m2)*s
+B3*B1*K+B3*B2*K)
>>
>> m1=3; m2=4; m3=5; B1=1; B2=2; B3=3; K=1000;
>> H=subs(H)
H =
(4*s^2+2*s+3000)/(60*s^4+122*s^3+9000+35054*s^2+57006*s)
>>
Using Maple to Find the Transfer Function:
The following page shows a Maple 11 classic work-sheet to solve the algebraic equa
tions. This example uses the procedures GenerateMatrix() and LinearSolve() from the
LinearAlgebra package.
Maple uses a : to end a command if no output is to be generated, or a; if the
output is to be displayed.
The LinearAlgebra package is loaded using the command with(LinearAlgebra):
Commands are grouped within execution groups indicated at the very left margin. A
group is executed by placing the cursor within a group and hitting Enter.
GenerateMatrix() requires that the equations and the output variables entered as
lists, that is [item1, item2, ...].
LinearSolve(A,b) solves linear equations in matrix form Ax = b where A, and b
have been generated from the equations using GenerateMatrix().
113
F
m
m m
B
B
B
K
1
1
2
2
3
3
!
"
m
v
1

Using Maple Syrep to Find the Transfer Function:
Maple Syrep ( for System Representation is a collection of approximately 100 functions for
symbolic and numerical analysis of dynamic systems. In this handout we examine two of its
functions for deriving a system directly from its linear graph or impedance graph.
We start using the graph for the system, but this time numbering each node:
The system is specied as a Maple list-of-lists, each contained in square brackets, for example
for the above diagram:
graph:=[[1,2,Force,Fs],
[2,1,mass,m3],
[2,1,damper,B3],
[2,3,damper,B2],
[2,4,damper,B1],
[3,1,mass,m2],
[4,1,mass,m1,vm1,Fm1],
[3,4,spring,K]]:
Each entry in the list represents a branch with the rst two items the nodes that it connects
to (in the direction of the arrow), the third describes the branch, the fourth the parameter,
and optionally two items naming the variables on the branch.
The system is generated using the procedures
LGraph to system(graph,output) - if the system is represented in linear graph form,
or
ZGraph to system(graph,output) - if the system is represented in impedance graph
form (where the elemental impedances are used).
Note: The newer versions of Maple allow a new mode of input, the 2D-Math Input mode that
allows formatting as you enter the data. With this mode turned on (it is now the default),
you can enter subscripts etc. Maple Syrep was written before this mode was available, and
uses the underscore ( ) in many function names, for example Transfer function(). In 2D
Math mode, this has to be entered as Transfer\_funtion() where the \ acts as an escape
character. I prefer to set Maple into the classic 1-D Math Input mode while using Syrep.
Examples of both modes are shown below.
114
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
+
-
v
v
v
out
+
-
Gai n: A

Spring Term 2008
Lecture 12
1
Reading:
Class Handout - Introduction to the Operational Amplier
Nise: pages 5558 (Operational Ampliers)
This lecture is inserted in the course at this point to coincide with Laboratory
Session 3, which requires the construction of a proportional controller using
a dierential amplier.
1 Introduction to the Operational Amplier
The integrated-circuit operational-amplier (op-amp) is a fundamental building block for
many electronic circuits, including analog control systems. In this hand-out we examine
some of the basic circuits that can be used to implement control systems. We take simplied
approaches to the analysis, and this discussion is by no means complete or exhaustive.
What is an operational amplier? It is simply a very high gain electronic amplier, with
a pair of dierential inputs. Its functionality comes about through the use of feedback around
the amplier, as we show below.
The op-amp has the following characteristics:
It is basically a three terminal amplier, with two inputs and an output. It is a
dierential amplier, that is the output is proportional to the dierence in the voltages
applied to the two inputs, with very high gain A,
v
out
= A(v
+
v
)
where A is typically 10
4
10
5
, and the two inputs are known as the non-inverting (v
+
)
and inverting (v ) inputs respectively. In the ideal op-amp we assume that the gain A
is innite.
1
121
1
2
3
4 5
6
7
8
-V
+V
output
offset bal ance
ss
ss
noni nverti ng i nput
i nverti ng i nput
offset bal ance
+
-
+
-
+ 1 5v
-1 5v
output i nput
+
-
v v
out
i n
In an ideal op-amp no current ows into either input, that is they are voltage-controlled
and have innite input resistance. In a practical op-amp the input current is in the
order of pico-amps (10
12
) amp, or less.
The output acts as a voltage source, that is it can be modeled as a Thevenin source
with a very low source resistance.
Op-amps come in many forms and with a bewildering array of specications. They range in
cost from a few cents to many dollars, depending on the specs. These specications include
input impedance, input bias current, output oset voltage, external power requirements, etc.
Higher grade ampliers are known as precision, or instrumentation ampliers.
Op-amps come in a variety of packages. A common inexpensive op-amp, the 741, has
an 8 pin DIP (dual in-line package) form. Many amps use a common basic pin-out for this
package as shown below:
The pins are numbered counter-clockwise from the top left as shown above. (Note that pin
1 is identied by a notch at the top or a dot beside pin 1.) The basic amplier is connected
between pins 2, 3 and 6. The amplier requires a pair of external supply voltages to operate,
these are typically 15 volts and are are connected to pins 7 (positive) and 4 (negative). Pins
1 and 5 are usually used for optional external oset nulling circuitry - the actual connection
is dependent on the type. We will not use this feature if we can get away without it.
Non-Inverting Congurations
The Unity Gain Non-Inverting amplier: The simplest conguration of the op-amp
is as a unity gain buer amplier:
The output is connected directly to the inverting input. Then
v
out
= A(v
+
v
)
= A(v
+
v
out
)
A
= v
+
1 + A
122
+
-
+ 1 5v
-1 5v
output i nput
+
-
v
v
out
i n
R
R
1
2
R R
2 1
+
-
Load
R
i
v
and if, as noted above, A is typically 10
4
or much higher we may simply say
v
out
= v
+
The unity gain amplier is used as a buer to minimize loading on a circuit, because it has a
very high input impedance (draws no current) yet provides a low output resistance output.
The Non-Inverting amplier with gain: The above circuit may be modied to feed
back only a fraction of the output voltage by including a two resistor voltage divider.
In this case R
1
and R
2
act as a voltage divider, and
R
2
v = v
out
R
1
+ R
2
Then
v
out
= A(v
+
v
in
)

R
2

= A v
in

R
1
+ R
2
v
out
A(R
1
+ R
2
)
= v
in
R
1
+ R
2
+ AR
2
and when A 1
R
1
+ R
2
v
out
= v
in
R
2
For example if R
1
= 47 k, and R
2
= 10 k, the amplier voltage gain will be 5.7.
The Voltage-Controlled Current-Source: The above circuit can be modied to pro
duce a current source.
123
+
-
+ 1 5v
-1 5v
output
+
-
v
v
out
i n
R
R
i n
f
i
i
i n
f
summi ng j uncti on
(vi rtual gound)
Gnd
i nput
R
R
i n
f
Assume that the op-amp drives a load of unknown resistance. A small resistor R is placed
in series with the load. Then the inverting input is v = iR. At the output of the amplier

v
out
=
=
A (v
+
v
)
A (v
in
iR)
Then
i =
1
R

v
in

1
A
v
out

and if A 1
i =
1
v
in
R
which is independent of the load. For example, if R = 0.1, then the current in the load
will be i = 10v
in
amps.
It should be realized, however, that typical op-amps are not capable of supplying more
than a few milli-amps of current. A power op-amp or an additional power amplier will
generally be necessary to supply signicant current.
Inverting Congurations
The Inverting Amplier: The basic inverting amplier has a conguration as shown
below:
The non-inverting input is connected to ground and a pair of resistors, R
in
and R
f
(the
feedback resistor) is used to dene the gain. To simplify the analysis we make the following
assumptions:
(a) The gain A is very high and we let it tend to innity. Because v
out
= A(v
+
v
) and
v
+
= 0, this means that v = lim
A
v
out
/A = 0. In other words, the inverting input
voltage is so small (volts) that we declare it to be a virtual ground, that is in our

simplied analyses we assume v = 0.

(b) No current ows into either of the inputs. This allows us to apply Kirchos current
law at the junction of the resistors.
With the above assumptions, when we apply Kirchos current law at the junction of the
resistors (the inverting input terminal):
i
in
+ i
f
= 0
124
+
-
+ 1 5v
-1 5v
output
+
-
v
v
out
R
f
i
f
summi ng j uncti on
(vi rtual gound)
Gnd
i nputs
R
R
f
R
R
1
2
1
2
v
i
i
1
2
R
1
2
Since the junction is at the virtual ground (v

0) we can simply use Ohms law:
v
in
v

+
V
out
v

v
in
+
V
out
= 0
R
in
R
f

R
in
R
f
or
R
f
v
out
= v
in
R
in
In other words the voltage gain is dened by the ratio of the two resistors. The term invert
ing amplier comes about because of the sign change.
The Inverting Summer: The inverting amplier may be extended to include multiple
inputs:
As before we assume that the inverting input is at a virtual ground (v

0) and apply
Kirchos current law at the summing junction
i
1
+ i
2
+ i
f
= 0
v
1
v
2
V
out
+ + = 0
R
1
R
2
R
f
or
R
f
R
f
v
out
=
R
1
v
1

R
2
v
2
which is a weighted sum of the inputs. If R
1
= R
2
= R
in
the output is proportional to the
sum of the input voltages.
R
f
v
out
=
R
in
(v
1
+ v
2
)
The summer may be extended to many inputs by simply adding input resistors R
i
n
v
out
=

R
R
f
i
v
i
i=1
If all of the input resistors are equal (R
i
= R
in
)
n
v
out
=
R
f

v
i
R
in
i=1
125
+
-
+ 1 5v
-1 5v
output
+
-
v
v
out
i n
R
i n
i
i
i n
f
summi ng j uncti on
(vi rtual gound)
Gnd
i nput
R
i n
C
C
+
-
+ 1 5v
-1 5v
output
+
-
v
v
out
i n
R
f
i
i
i n
f
summi ng j uncti on
(vi rtual gound)
Gnd
i nput
R
f
C
C
The Integrator: If the feedback resistor in the inverting amplier is replaced by a capac
itor C the amplier becomes an integrator.
At the summing junction we apply Kirchos current law as before but the feedback current
is now dened by the elemental relationship for the capacitor:
i
in
+ i
f
= 0
v
in
+ C
dv
out
= 0
R
in
dt
Then
dv
out
1
dt
=
R
in
C
v
in
or
v
out
=
1
R
in
C

t
0
v
in
dt + v
out
(0)
As above, the integrator may be extended to a summing conguration by simply adding
extra input resistors:
t
n
1

v
i

dt + v
out
(0) v
out
C
=
0
i=1
R
i
and if all input resistors have the same value R
n
1

t

v
out
= v
i
dt + v
out
(0)
RC
0
i=1
The Dierentiator: If the input resistor R
in
in the inverting amplier is replaced by a
capacitor, the circuit becomes a dierentiator:
126
+
-
v
v
out
i n
i
i
i n
f
summi ng j uncti on
(vi rtual gound)
Z
Z
f
i n
(s)
(s)
+
-
v
v
out
R
i
i
i n
f
1
1
2
2
R R
v
1
2
At the summing junction we apply Kirchos current law as before but the input current is
now dened by the elemental relationship for the capacitor:
i
in
+ i
f
= 0
dv
in
v
out
C + + = 0
dt R
f
Then
1 dv
in
v
out
=
R
in
C dt
Pure dierentiators are rarely used because of their susceptibility to noise. In practice it is
common to use a pseudo-dierentiator by including a resistor in series with the capacitor.
A General Linear Transfer Function: The inverting conguration may be extended to
synthesize many general linear transfer functions by including appropriate networks in the
feedback and input paths. Consider the general circuit shown below with elements described
by impedances Z
in
(s) and Z
f
(s) in the input and output paths:
At the summing junction
i
in
+ i
f
= 0
V
in
(s) V
out
(s)
+ + = 0
Z
in
(s) Z
f
(s)
so that the amplier becomes a linear system with a transfer function
V
out
(s) Z
f
(s)
H(s) = =
V
in
(s) Z
in
(s)
This conguration may be used to synthesize lead-lag compensators for control, active lters
for signal processing, and many other functions.
The Dierential Amplier
A dierential amplier produces an output that is proportional to the dierence between
two inputs.
127
At the non-inverting input
R
2
v
+
= v
1
R
1
+ R
2
If the op-amp gain is assumed to be innite, v = v
+
. The input current at the inverting
amplier is:
i
in
=
v
2
v

=
v
2
v
+
R
1
R
1
1 R
2
= v
1
R
1
v
2

R
1
(R
1
+ R
2
)
Similarly the feedback current is
i
f
=
v
out
v

=
v
out
v
+
R
2
R
2
1 1
=
R
2
v
out

(R
1
+ R
2
)
v
1
Applying Kirchos current law at the summing junction i
in
+ i
f
= 0, then yields
R
2
v
out
= (v
1
v
2
)
R
1
so that the output is proportional to the voltage dierence between the two inputs.
Practical Considerations:
The above simplied description does not include many practical considerations that must
be taken into account when designing high performance op-amp circuits, including:
Most op-amps are low current devices. Most circuits are designed to operate with
signal currents in the mA or A range. This means that resistor values should be in
the range 10 k and above.
An op-amp cannot supply voltage levels greater than the supply voltage levels. Care
must be taken to ensure that the output voltage remains within the limits dened by
the supply voltage.
Our simplied analysis has not taken bandwidth limitations into account. Practical op
amps exhibit a high-frequency roll-o of gain because of internal capacitances within
the chip.
A similar phenomenon is know as slew-rate limiting. In practice, op-amps have a
limit on the slope (v/s) of the output voltage. This nonlinear eect has a limiting
eect on the high frequency performance.
128
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008

Spring Term 2008
Lecture 13
1
Reading:
Class Handout: Modeling Part 2 Summary of One-Port Primitive Elements
1 Generalized System Representation
In the electrical and mechanical systems we have dened, and used, two power variables
(variables whose product is power):
Electrical
P = vi, where
v voltage drop
i current
Mechanical
P = Fv, where
v velocity drop
F force
We have also developed a common graph notation to express the system topology, and have
used extensions to Kirchos laws to draw analogies between pairs of system variables:
(1) voltage drop and velocity drop, and
(2) current and force.
In addition, we drew analogies between the primitive modeling elements in each domain:
(1) capacitors and masses
(2) inductors and springs
(3) resistors and dampers.
We now look at generalizing these associations.
Variables: In each energy domain we will class the two power variables as either:
(a) Those variables that are dened by measuring a dierence, or drop, across an element,
that is between nodes on a graph (across one or more branches). These variables sum
to zero around any closed loop on the graph (they satisfy the compatibility conditions).
These variables are dened to be across-variables.
The two across variables we have dened so far are (i) velocity drop in mechanical
systems, and (ii) voltage drop in electrical systems.
Example: To measure voltage drop in an electrical circuit you must connect a volt
meter across an element:
1
131
A
R
i
v
R
+
-
An ammeter measures current
"through" an el ement.
V
F(t)
m
K
F
Force i s measured by i nserti ng an i nstrument i n seri es
wi th an el ement.
V
R v
R
+
-
A vol tmeter measures vol tage
"across" an el ement
V
R
1
2
2
(b) Those variables that are measured through an element, that is are considered as being
transmitted through an element unchanged. These variables sum to zero at the nodes
on a graph, and are said to satisfy the continuity condition. They are dened to be
through variables.
The through variables dened so far are (i) current in electrical systems, and (ii) force
in mechanical systems.
Example: to measure force in a mechanical system, or current in an electrical system
a sensor mus be inserted in series with an element.
Generalized Variables: In dealing with generalized modeling power variables, without
regard for a particular energy domain we now dene
A generalized across variable, designated v, and
A generalized through variable, designated f.
We also generalize the constraints on across and through variables imposed by the structure
of a systems graph:
The compatibility condition
n

v
i
= 0 around the n elements in any loop on a graph
i=1
which is clearly a generalization of Kirchos voltage law, and
The continuity condition
n

f
i
= 0 in the n elements connected to any node on a graph
i=1
which is clearly a generalization of Kirchos current law.
132
On an across-vari abl e source the arrow poi nts
i n the di recti on of the assumed across-vari abl e
drop, that i s we assume
v > v
V
v
v
>
=
=
>
On an through-vari abl e source the arrow poi nts
i n the di recti on of the assumed through-vari abl e
"fl ow".
.
+
L
L
= >
L
+
E
Generalized Sources: Along with the generalized variables we dene a pair of ideal sources
(1) The Across Variable Source: An across variable source maintains the across vari
able between the nodes it is connected to, regardless of the magnitude of the through
variable supplied to the system.
The examples we have seen so far are the voltage source in electrical systems, and the
velocity source in mechanical systems.
(2) The Through Variable Source: An through variable source maintains the through
variable into the nodes it is connected to, regardless of the magnitude of the across
variable it must generate to maintain its through variable.
The examples we have seen so far are the current source in electrical systems, and the
force source in mechanical systems.
Generalized Elements: We have seen that in each of the energy domains studied so
far we have dened three primitive modeling elements: two energy storage elements and
a dissipative element. We now classify these elements according to (i) the variable that
accounts for the stored energy in energy storage elements, and (ii) we group together the
dissipative elements.
A-Type Elements These are the energy storage elements in which the stored energy is a
function of the across-variable.
Electrical: For a capacitor, let v = v
a
v
b
be the across variable:
dv
i = C
dt
E =

t
vi dt =

t
Cv dv
=
1
Cv
2
0
2
133
F m
v
m
v
ref
f = +
@v
@J
general i zed through vari abl e
general i zed capaci tance
general i zed across vari abl e
L
v v
i
a
b
a b
v = v - v
L
which denes the capacitor as an A-type element.
Mechanical: For a mass element
dv
F = m
dt
E =

t
vF dt =

t
mv dv
=
1
mv
2
0
2
which denes the mass element as an A-type element.
Dene a generalized capacitance C that represents A-type elements inde
pendent of the energy domain, and write its elemental equation in terms of
generalized across and through variables:
A-Type elements may be summarized as in the following table:
Element Elemental equation Energy
Generalized A-type f = C
dv
dt
E =
1
2
Cv
2
Translational mass
Electrical capacitance
F = m
dv
dt
i = C
dv
dt
E
E
=
=
1
2
mv
2
1
2
Cv
2
T-Type Elements These are the energy storage elements in which the stored energy is a
function of the through-variable.
Electrical: For a inductor, let v = v
a
v
b
di
v = L
dt
E =

t
vi dt =

t
Li di
=
1
Li
2
0
2
134
a
K
F F
v
v a
b
b
v = v - v
K
v = L
@B
@J
general i zed i nductance
L
=
4
E
L
>
= >
v = v - v
4
which denes the inductor as an T-type element.
Mechanical: For a spring element
1 dF
v =
K dt
E =

t
vF dt =
1

t
F dF
=
1
F
2
K
0
2K
which denes the spring element as an T-type element.
Dene a generalized inductance L that represents T-type elements indepen
dent of the energy domain, and write its elemental equation in terms of
T-Type elements may be summarized as in the following table:
Generalized T-type v = Ldf/dt E =
1
2
Lf
2
Translational spring
Electrical inductance
v =
1
K
dF
dt
v = L
di
dt
E =
1
2K
F
2
E =
1
2
Li
2
D-Type Elements These are the dissipative elements (non-energy storage, the power
ow is always into the element).
Electrical: For a resistor, let v = v
a
v
b
v = iR
P = vi = i
2
R = v
2
/R
0
135
. .
*
L
=
L
>
= >
v = v - v
*
v = 4f
general i zed resi stance
which denes the resistor as an D-type element.
Mechanical: For a damper element
F = Bv
P = vF = Bv
2
= F
2
/B
0
which denes the damper element as an D-type element.
Dene a generalized resistance R that represents D-type elements indepen
dent of the energy domain, and write its elemental equation in terms of
D-Type elements may be summarized as in the following table:
Element Elemental equations Power dissipated
Generalized D-type f =
1
R
v v = Rf P =
1
R
v
2
= Rf
2
Translational damper
Electrical resistance
F = Bv
i =
1
R
v
v =
1
B
F
v = Ri
P = Bv
2
=
1
B
F
2
P =
1
R
v
2
= Ri
2
Generalized Impedances: The generalized impedance of an is
V (s)
Z =
F (s)
where V (s) and F (s) are the Laplace domain representation of the generalized across and
through variables. The impedances are summarized in the following table:
A-Type T-Type D-Type
Generalized
1
Cs
sL R
Translational
1 1
s
1
sm K B
1
Electrical
Cs
sL R
136
C
s
R
I m
K
B
v
m
F
v = 0
one i dependent energy storage el ement two i dependent energy storage el ements
H(s) =
1
RCs + 1
(order = 1 ) H(s) =
1
ms + Bs + K
(order = 2)
2
m
K K
1
2
v
F
K and K are not i ndependent energy storage
el ements si nce E E .
Thi s i s a second-order system.
K K
1
2
1
1
2
F m
K
2
K
1
K and K are effecti vel y i n paral l el and
may be repl aced by an equi val ent spri ng.
1 2
m
m
1
v
v
1
2
2
m and m are not i ndependent energy storage
el ements because v v and therefore thei r
energi es are al gebrai cal l y rel ated.
2
2
1
1
2 Notes on Transfer Function Generation
The system order (highest order derivative on the l.h.s. of the dierential equation, or the
highest power in s in the denominator of the transfer function) is determined by the number
of independent energy storage elements in the system.
An independent energy storage element is one whose stored energy is not directly propor
tional to the energy stored in other elements or completely dened by the system sources.
For example, in the following gure the two springs are not independent energy storage ele
ments, because their stored energies are directly related. Therefore this system will generate
a second-order dierential equation even though there are three energy storage elements.
similarly, in the following gure the velocities of the two masses are related by the lever, and
they are not independent energy storage elements.
Another situation where an energy storage element does not increase the order of a system
is when the stored energy is completely dened by a source.
(a) Any element connected directly in parallel with an across variable source will not
aect the system dynamics, and in fact will not appear in the dierential equation (or
137
m
K
v
V
m
1
2
m
2
vel oci ty
source
m
m
m
K K
1
2
2
V
The vel oci ty of m i s compl etel y defi ned by the source
1
V
C
R L
R
1
2
I
I
R
2
R
1 L
C
I
R
2
C
I
R and L do not affect the system dynami cs si nce the
source defi nes the current through them.
1
transfer function). The reason is that the source completely denes the across variable
at its two nodes, and therefore the element can be removed without aecting the rest
of the system. For example in the following system the output is the velocity of the
mass m
2
. The system input is a velocity source V applied to the mass m
1
. The velocity
at the left-hand node is completely dened by the source, and would not be aected if
m
1
was removed from the system. The transfer function relating v
m
2
to V is
V
m
2
(s) K/m
2
H(s) = =
V (s) s
2
+ K/m
2
and does not involve m
1
.
(b) Any element connected directly in series with a through variable source will not aect
the system dynamics. The through variable transmitted from the source to the rest of
the system will be unchanged by the presence of the series element(s). In the electrical
system below the resistance R
1
and the inductor L are in series with the current
source I. Since i
L
= I, the current supplied to the right-hand node is unaected by
the presence if either L or R
1
, and this is eectively a rst-order system.
138
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
W, 6

Spring Term 2008
Lecture 14
1
Reading:
Sec. 3.
Class Handout: Modeling Part 2: Summary of One-Port Primitive Elements
1 The Modeling of Rotational Systems.
With the the modeling framework as we dened it in Lecture 13, we have seen that in each
energy domain we need to dene
(a) Two power variables, an across variable, and a through variable. the product of these
variables is power.
(b) Two ideal sources, and across variable source, and a through variable source.
(c) Three ideal modeling elements, two energy storage elements (a T-type element, and a
A-Type element), and a dissipative (D-Type) element.)
(d) A pair of interconnection laws.
We now address modeling of rotational mechanical systems.
(a) Denition of Power variables: In a rotational system we consider the motion of a
system around an axis of rotation:
Consider the rotary motion resulting from a force F applied at a radius r from the
rotational axis
1
141
W , T
W , T
1
2
Angul ar vel oci ti es W and W can be
di fferent across an el ement.
The torque T transmi tted through an
el ement i s constant.
1 2
.
H
,N
G
9
The work done by the force F in moving an innitesimal distance x is
W = F x = Fr
and the power P is
dW d
P = = F r = T
dt dt
where T = Fr is the applied torque (N.m), and = d/ dt is the angular velocity
(rad/s).
We note that if T and have the same sign, then P > 0 and power is owing into
the system or element that is being rotated. Similarly, if T and have the opposite
signs, then P < 0 and power is owing from the system or element, in other words the
system is doing work on the source.
Note that the angular velocity can be dierent across an element, but that torque
T is transmitted through an element:
We therefore dene our power variables as torque T and angular velocity , where
T is chosen as the through variable
is chosen as the across variable.
(b) Ideal Sources: With the choice of modeling variables we can dene our pair of ideal
sources
The Angular Velocity Source:
s
(t)
By denition the angular velocity source is an across variable source.The ideal angular
velocity source will maintain the rotational speed regardless of the torque it must
generate to do so:
142
6
9
6
torque
Torque 6 i s i ndependent
of angul ar vel oci ty 9
arrow shows di recti on

of torque.
r
9
m
W
W
6
torque
angul ar vel oci ty W i s
i ndependent of torque 6
arrow shows di recti on

of assumed ang. vel .
drop.
W
W
=
>
W > W
= >
The Torque Source: T
s
(t)
By denition the torque source is a through variable source. The ideal torque source
will maintain the applied torque regardless of the angular velocity it must generate to
do so:
(c) Ideal Modeling Elements:
1 The Moment of Inertia: Consider a mass element m rotating at a xed radius
R about the axis of rotation.
The stored energy is
1 1
E = m(r)
2
= J
2
2 2
where J = mr
2
is dened to be the mo
ment of inertia of the particle.
For a collection of n mass particles m
i
at radii r
i
, i = 1, . . . , n, the moment of inertia
is
n
2
J =

m
i
r
i
.
i=1
143
r
mass dm at
radi us r
9
L
9
J =
mL
1 2
2
where m i s the mass
of the rod
r
where m i s the mass
of the di sk
J =
1
2
2
mr
For a continuous distribution of mass about the axis of rotation, the moment of inertia
is
J =

R
r
2
dm
0
Examples:
A uniform rod of length L
rotating about its center.
A uniform disc with radius r
rotating about its center.
The elemental equation for the moment of inertia J is
d
J
T
J
= J
dt
We note that the energy stored in a rotating mass is E = J
2
/2, that is it is a function
of the across variable, dening the moment of inertia as an A-type element.
As in the case of a translational mass element, the angular velocity drop associated
with a rotary inertia J is always measured with respect to a non-accelerating reference
frame.
Elemental Impedance: By denition
J
(s) 1
Z
J
= =
T
J
(s) Js
from the elemental equation.
144
9
9
shaft can twi st wi th
di fferent angul ar
vel oci ti es at the
two ends
1
2
J
J
1 2
9
6
coi l spri ng
(2) The Torsional Spring:
W , q
a
K
T
a
b b
T
W , q
W = W - W
b a
Let
a
and
b
be the angular displacements of the two ends from their rest positions.
Hookes law for a torsional spring is
T = K(
a

b
).
where K is dened to be the torsional stiness. Dierentiation gives
dT d(
a

b
)
= K
dt dt
dT
= K
dt
where = (

a

b
) is the angular velocity drop across the spring.
Torsional stiness may result from the material properties of a long shaft
or may be intentional, for example in a coil (hair) spring in a mechanical watch.
145
W
=
6
>
6
W
vi scous fri cti onal
contact
W = W - W
> =
=
6
9
external vi scous
fri cti onal contact
beari ngs
(fri cti on)
shaft
6, 9
The energy stored in a torsional spring is
t

1
E = T dt = T
2
2K

which is a function of the through variable, dening the spring as a T-type element.
K
(s) s
Z
K
= =
T
K
(s) K
(3) The Rotational Damper: We look for an algebraic relationship between T and
of the form
T = B
which is approximated as viscous rotational friction:
Notice that P = T = B
2
> 0, which denes the damper as a D-type element.
B
(s) 1
Z
B
= =
T
B
(s) B
(d) Interconnection Laws: Consider an inertial element J subject to n external torques
T
1
, T
2
, . . . , T
n
, for example
146
1
2
3
J
9 = 0
ref
9
J
arrow on an
i nerti al branch
al ways poi nts to
the reference node
9 i s al ways measured
wi th repect to an i nerti al
reference frame
J
1
T
9
9 K
B
2
9
a
K
B
=

4
J
T T T T
1 2
3
9
reference
di recti on
then
d
J = T
1
T 2 + T
3
+ T
4
dt
and in general
n
d
T
i
= J
dt
i=1
As in the translational case, we consider a ctitious dAlembert torque T
j
and write
n
T
i
T
J
= 0

i=1
as the torque balance (continuity condition) at a node.
For an inertia-less node (J = 0),
n
T
i
= 0
i=1
which states that the external torques sum to zero, for example at node (a) below,
T
B
T
K
= 0.
147
(a)
(b)
J
B
K
9 = 0
1
2
a
9
9
K
b
J
B
Continuity Condition: The sum of torques (including a dAlembert torque
associated with an inertia a element) at any node on a system graph is zero.
Nodes represent points of distinct angular velocity in a rotational system, and by
analogy with translational systems, the compatibility condition is
Compatibility Condition: The sum of angular velocity drops around any
closed loop on a system graph is zero.
For example, on the graph:
two compatibility equations are:
K
+
J

s
= 0 (Loop 1),
B

J
= 0 (Loop 2).
2 Updated Tables of Generalized Elements to Include Rotational
Elements:
The tables presented in Lecture 13 are now updated to include rotational systems.
A-Type Elements:
Generalized A-type f = C
dv
dt
E =
1
2
Cv
2
Translational mass
Rotational inertia
Electrical capacitance
F = m
dv
dt
T = J
d
dt
i = C
dv
dt
E
E
E
=
=
=
1
2
mv
2
1
2
J
2
1
2
Cv
2
148
T-Type Elements :
Generalized T-type v = Ldf/dt E =
1
2
Lf
2
Translational spring
Torsional spring
Electrical inductance
v =
1
K
dF
dt
=
1
K
dT
dt
v = L
di
dt
E =
1
2K
F
2
E =
1
2K
T
2
E =
1
2
Li
2
D-Type Elements:
Element Elemental equations Power dissipated
Generalized D-type f =
1
R
v v = Rf P =
1
R
v
2
= Rf
2
Translational damper
Rotational damper
Electrical resistance
F = Bv
T = B
i =
1
R
v
v =
1
B
F
=
1
B
T
v = Ri
P = Bv
2
=
1
B
F
2
P = B
2
=
1
B
T
2
P =
1
R
v
2
= Ri
2
Generalized Impedances:
A-Type T-Type D-Type
Generalized
1
Cs
sL R
Translational
1 1
s
1
sm K B
1 1 1
Rotational s
sJ K B
1
Electrical
Cs
sL R
149
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
in
K
9
G
J
B
J
9
fri cti onal drag pl ates

Spring Term 2008
Lecture 15
1
Reading:
Sec. 3.
Class Handout: Modeling Part 2: Summary of One-Port Primitive Elements
Nise: Secs. 2.4 and 2.6.
1 Rotational Systems (continued)
Example 1
The diagram shows a mechanical tachometer that uses frictional drag plates to
create a torque proportional to angular velocity dierence. The angular velocity
is indicated by the displacement of a torsional spring.
Find the transfer function relating the displacement of the indicator to the
input angular velocity
in
(s)
H(s) =
in
(s)
and show that for a constant input angular velocity the steady-state indicated
speed
ss

i
.
Solution: There are two distinct angular velocities, and the system graph is:
1
151
J
B
K
9 = 0
in 9
across-vari abl e
source
Using impedances, redraw the graph combining the inertia and the spring into a
single impedance Z
2
:
Z
Z
9 = 0
in 9
J
9
1
2
T
Z =
1
1
B
1
Z =
2
1
Js
s
K
=
1/JK
s/K + 1/Js
Then
Z
2
J
=
in
(s)
Z
1
+ Z
2
s/(Js
2
+ K)
=
in
(s)
1/B + s/(Js
2
+ K)
Bs
=
in
(s).
Js
2
+ Bs + K
But the angular displacement (s) =
J
(s)/s so that
B
H(s) =
Js
2
+ Bs + K
If the input velocity is a step function at t = 0, the
in
(s) = /s and the
steady-state indicated value will be
B
ss
= lim (t) = lim s(s) = lim sH(s) =
t s0 s0
s K
show that the indicated speed () is proportional to the input angular velocity.
(This example is repeated using mesh equations below.)
152
m
m
B
B
B
1
1
2
2
3
K
1
2
Vt
K
v
o
1
K
K
2
m
1
m
2
B
1
B
2
B
3
Vt
v
o
1
Z
Z
5
Z
2
Z
4
Z
3
V(t)
v
o
v = 0
s
K
Z =
3
1
B
3
Z =
2
1
B
1
1
m s
1
Z =
4
1
B
2
1
m s
2
Z =
5
s
K
2
Z =
1
1
1
Z
Z
5
Z
2
Z
4
Z
3
Vt
v
v = 0
o
1
2
3
2 Transfer Function Generation Using Mesh/Loop Equations
This method is useful when the system contains an across variable source.
We will develop the method using the following mechanical system as an example.
The input is a velocity source V (t), and the output is the velocity of the mass m
2
.
Step 1: (This step is optional.) Do some simple impedance combinations to reduce the
complexity, being careful not to eliminate any nodes or branches that specify the output
variable.
Step 2: Dene a set of closed loops/meshes, making sure that each branch is covered by
at least one loop.
Step 3: Write a loop equation for each loop:
v
Z
1
+ v
Z
2
V = 0
v
Z
3
+ v
Z
4
v
Z
2
= 0
v
Z
5
v
Z
3
v
Z
1
= 0
153
in
K
9
G
J
B
J
9
fri cti onal drag pl ates
J
B
K
9 = 0
in 9
1
2
Step 4: At this point we assume that each loop has a continuous through variable with
it, that is loop n has F
n
associated with it. When a branch is contacted by more
that one loop, the branch through variable is the sum of all of the contacting loop
through variables. We use impedance relationships and substitute loop through variable
expressions for the across variables:
Z
1
(F
1
F
3
) + Z
2
(F
1
F
2
) V = 0
Z
3
(F
2
F
3
) + Z
4
F
2
Z
2
(F
1
F
2
) = 0
Z
5
F
3
Z
3
(F
2
F
3
) Z
1
(F
1
F
3
) = 0
Step 5: Rewrite these equations collecting terms in the loop through variables to create a
set of linear equations:
(Z
1
+ Z
2
)F
1
Z
2
F
2
Z
1
F
3
= V
Z
2
F
1
+ (Z
2
+ Z
3
+ Z
4
)F
2
Z
3
F
3
= 0
Z
1
F
1
+ Z
3
F
2
+ (Z
1
+ Z
3
+ Z
5
)F
3
= 0
It may be useful to express these equations in matrix form:
(Z
1
+ Z
2
) Z
2
Z
1
(Z
2
+ Z
3
+ Z
4
) Z
3
F
1
F
2
V
0 Z
2
Z
1
Z
3
(Z
1
+ Z
3
+ Z
5
) F
3
0
Step 6: Identify how the output variable is related to the loop through variables. In this
case
v
out
= F
2
Z
4
Step 7: Solve the set of linear equations for the variable(s) identied in Step 6, in this
case F
2
, using any method, and substitute in the output equation (Step 6).
(we will not do this step here).
Example 2
Repeat the mechanical tachometer problem of Example 1, this time using mesh
equations.
154
With the system graph and two loops dened as above, the loop equations are:
B
+
J

in
= 0
K

J
= 0
Substitute for the across variables ():
Z
B
T
1
+ Z
K
(T
1
T
2
) =
in
Z
J
T
2
Z
K
(T
1
T
2
) = 0
Rearrange:
(Z
B
+ Z
K
)T
1
Z
K
T
2
=
in
Z
K
(T
1
+ (Z
J
+ Z
K
)T
2
= 0
and in matrix form

Z
B
+ Z
K
Z
K

T
1

in

=
Z
K
Z
J
+ Z
K
T
2
0
The output variable is
J
= Z
J
T
2
, therefore use Cramers rule to solve for T
2
.
Z
B
+ Z
K
in
Z
K
0 Z
K
in
T
2
=
Z

B
+ Z
K
=
Z
B
Z
J
+ Z
B
Z
K
+ Z
K
Z
J
Z
K
Z
K
Z
J
+ Z
K
and since
J
= Z
J
T
2
, Z
J
= 1/(Js), Z
K
= s/K, and Z
B
= 1/B
Z
J
Z
K
in
Bs
J
= =
in
Z
B
Z
J
+ Z
B
Z
K
+ Z
K
Z
J
Js
2
+ Bs + K
and as before, (s) =
J
/s giving
(s) B
H(s) = =
in
(s) Js
2
+ Bs + K
3 Transfer Function Generation Using Node Equations
This method is useful when the system contains a through variable source.
We will demonstrate the method using the following graph:
155
K
B
v
m m
1
1
v
2
2
v = 0
F
s
K
?
1
>
=
m
m
2
B
Z
Z
Z
?
s
F
Z
1
2
3
4
>
=
Step 1: Write node equations for each node (except the reference node) - let f be a
generalized through variable.
F
s
f
Z
1
f
Z
2
= 0 node(a)
f
Z
2
f
Z
4
f
Z
3
= 0 node(b)
Step 2: Express the branch through variables in terms of admittances (Y = 1/Z) and
across variables
(v
a
v
b
)Y
2
+ (v
a
v
c
)Y
1
= F
s
(v
a
v
b
)Y
2
+ (v
b
v
c
)Y
4
(v
b
v
c
)Y
3
= 0
Step 3: Collect terms in the across variables, and form a set of linear equations (note that
node (c) is the reference node, therefore v
c
= 0).
(Y
1
+ Y
2
)v
a
Y
2
v
b
= F
s
Y
2
v
a
(Y
2
+ Y
3
+ Y
4
)v
b
= 0
Step 4: Identify the output variable in terms of nodal across variables and solve the linear
equations.
Example 3
Find H(s) = X
m
2
(s)/F (s) for the following system:
The graph has been simplied by combining parallel elements:
156
F
s
Z
?
1
>
=
Z
2
Z
3
Y =
1
Z
1
1
= m s
1
Y =
1
Z
2
2
= m s
2
Y =
1
Z
3
3
= B +
K
s
At the nodes (a) and (b):
F
B
F
m
1
F
s
= 0
F
B
F
m
2
+ F
s
= 0
Substitute for through variables
(v
a
v
b
)Y
3
v
a
Y
1
= F
s
(v
a
v
b
)Y
3
v
b
Y
1
= F
s
Reorganize and write in matrix form:

(Y
3
+ Y
1
) Y
3

v
a

F
s

=
Y 3 (Y
2
+ Y
3
) v
b
F (s)
Recognize that the output x
m
2
(s) = v
b
(s)/s, use Cramers rule to solve for v
b
:
F (s)
(Y
3
+ Y
1
)
Y 3 F
s
)
Y
1
F
s
v
b
= =
(Y
3
+ Y
1
) Y
3
Y
1
Y
2
+ Y
3
(Y
1
+ Y
2
)
Y 3 (Y
2
+ Y
3
)
Y
1
= m
1
s, Y
2
= m
2
s, and Y
3
= B + K/s, giving
m
1
s
v
m
2
(s) = v
b
(s) = F
s
(s)
m
1
m
2
s
2
+ B(m
1
+ m
2
)s + K(m
1
+ m
2
)
and
x
m
2
(s) 1 v
m
2
(s) m
1
H(s) = = =
F
s
(s) s F
s
(s) m
1
m
2
s
2
+ B(m
1
+ m
2
)s + K(m
1
+ m
2
)
157
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
L
L
8
=
> I
i ndi cates that we
assume L > L
>
=
=
>
.
I
i ndi cates that we assume the current
fl ow i n the di recti on of the arrow
or
that the source acts to move node =
i n the posi ti ve reference di recti on i n a
mechani cal system

Spring Term 2008
Lecture 16
1
Reading:
Class Handout - Modeling Part 3: Two-Port Energy Transducing Elements
1 Arrow Conventions on Ideal Sources
To this point we have simply told you to draw the arrows on source elements
(a) In the direction of the assumed across-variable drop for across-variable sources (voltage
and velocities),
(b) in the direction of the assumed through-variable direction for through-variable sources
(currents and forces)
When we draw a branch on a graph we make the assumption that P > 0, that is that
power is owing into the element. There are in fact two arrows implicit on each branch:
one representing the assumed across-variable drop, and a second representing the assumed
through-variable direction. If P > 0 (power is owing into the element), the two arrows are
in the same direction. For example, consider a capacitor
1
161
System
1
I
across-vari abl e
drop
through-vari abl e
di recti on
4
drop
+
-
E
4
drop
E
Z
Z
Z
v = 0
s
V
1
2
F
s
1
2
3
Loop 1 : v + v - V = 0
Z
1
s Z
2
Loop 2: v - v = 0
Z
3
Z
2
+
L
>
L
=
E
+
drop
L
+
=
>
E
+
L
+
=
>
vol tage drop
and current
i n the same
di recti on
vol tage drop
and current
i n the opposi te
di recti on
P = v
c
i
c
> 0 when either P = v
c
i
c
< 0 when either
1. v
c
> 0 and i
c
> 0,or 1. v
c
> 0 and i
c
< 0,or
On
2. v
c
< 0 and i
c
< 0. 2. v
c
< 0 and i
c
> 0.
passive elements, with the assumption P > 0 we can combine the two arrows into one because
they point in the same direction.
For sources, the assumption is the opposite - that is we assume that the source is supplying
energy/power to the system, P < 0.
There are two arrows (pointing in opposite directions) associated with any source
one for the across-variable, the second for the through-variable.
In modeling sources, we choose to show the arrow that will normally be used to solve the
system;
(a) An across-variable source will usually be included in a loop-equation, therefore the
convention is to show the arrow associated with the across-variable drop.
(b) A through-variable source will be included in a node-equation, therefore the convention
is to show the arrow representing the direction of the assumed through variable.
162
J
9
dri ve
motor
T
s
v
+
-
i
el ectri cal
rotati onal
a dc servo motor i s
a transducer between thee
el ectri cal and rotati onal
energy domai ns
l
l
v
v
1
1
2
2
F
1
2
F
a lever i s a transducer between
two transl ati onal domai ns (for
smal l di spl acements)
s
F
(a)
v = 0
Z
1 Z
2
at node (a): F - F - F = 0
Z
1
Z2 s
across-vari abl e
drop
opposi te to
di recti on of F
s
2 Energy Transduction Two-Port Elements
Reading: Class Handout - Modeling Part 3: Two-Port Energy Transducing Ele
ments
Many systems involve two or more energy domains, for example a system containing a dc
motor
or there may be a scaling of the across- and through-variables within a single domain, for
example a mechanical lever
The following two pages show some examples of two-port elements.
163
T
+
-
v
i
9
Area A v
F Q
P
Q
Q
T
9
9
P
P
Reservoi r
i = - T
1
K
__
v
(b) El ectri cal motor/generator
V = r9
(a) Rack and pi ni on
F = AP
(d) Fl ui d pi ston-cyl i nder
T = DP
(c) Rotary posi ti ve di spl acement pump
9 = - Q
1
D
_
r
T
v
F
v
F
9
9
Pi ni on
Rack
v = - Q
1
A
_
v = K 9
v
F = - T
1
r
_
9
F
v
F
v
T
9
v = -r9
F = T
1
r
_
r
+
-
i
F
magnet
coi l
E
v
E = K v
i = - F
1
K
__
(b) Sl i der-crank
(d) Movi ng-coi l l oudspeaker
164
F
v
F
v
1
1
2
2
l
1
l
2
+
-
+
-
v
v
i
i
N turns
N turns
1
2
1
1
2
2
v = v
2 1
1_
N
N = N / N
2 1
For a. c. i nputs
(e) El ectri cal transformer
Area: A
Area: A
P
P
Q
Q
P = 0
1
1
1
o
2
2
2
A = A /A
1 2
P = A P
1 2
1
2
Q = Q
1
A
_
-
(f) Fl ui d transformer
N teeth
N teeth
T T
9
9
9 9
2 1
1
1
2
2
1
2
N = N /N
2 1
9 = 9
1
N
_
-
1 2
T = N T
1 2
(b) Gear trai n
L = l /l
2 1
v = v
1 2
1
L
_
-
F = L F
1 2
(c) Mechani cal l ever
F
v F
v
1 1
2 2
T
1 T
9
2
2
9
2
9
1
r
2
1
r
9
1
T = 6
1 2
9 = R9
1 2
1
R
_
-
R = r /r
1 2
F = F
1 2
1
2
_
v = -2 v
1 2
(a) Bl ock and tackl e
(d) Bel t dri ve
i = - N i
1 2
165
TWO-PORT
TRANSDUCER
f
f
v
+
- -
+
P = f v
P = f v
transformer
f
v
gyrator
f
v
In all of these examples the energy transduction is lossless and static, that is there is no
energy storage.
P
1
+ P
2
= f
1
v
1
+ f
2
v
2
= 0
where the power ow at each port is dened to be positive into the port.
There are two possibilities:
(a) There is an proportional relationship between the across-variables on the two sides of
the two-port element, and an proportional relationship between the through-variables.
v
1
= kv
2
across-variable
1
across-variable
2
f
1
= (1/k)f
2
through-variable
1
through-variable
2
,
where clearly P
1
= P
2
. This relationship denes a transformer.
(b) There is an proportional relationship between the across-variables on one side and the
through-variable on the other side of the two-port element.
v
1
= kf
2
across-variable
1
through-variable
2
f
1
= (1/k)v
2
through-variable
1
across-variable
2
,
where clearly P
1
= P
2
. This relationship denes a gyrator.
166
L R
el ectri cal
J
B
e
b
+
+
-
-
rotati onal
V (t)
s
f
f
rotor
9
R
B
J V (t)
s
L
1
2
v = 0
9 = 0
i nherent resi stance and i nductance of
motor wi ndi ng
i deal transducer
Examples:
(1) Rack and Pinion:
V = r9
r
v
F
9
Pi ni on
Rack
F = - T
1
r
_
a transformer
}
It can be seen that the linear velocity of the rack is proportional to the angular velocity
of the pinion. Similarly the force needed to balance the torque applied to the pinion
is proportional to the torque. The rack and pinion is therefore a lossless transformer.
(2) DC motor:
9
dri ve
motor
T
v
+
-
i
T = -Ki
9 = v
1
K
a transformer
}
In the dc motor the torque produced is proportional to the current owing, while the
back emf produced is proportional to the angular velocity of the shaft. The motor is a
lossless transformer.
Example 1
A DC motor with an inertial load
167
Z = ? Z
1
2
= >
v = kv
1 2
f = - f
1 2
1
k
1
J
gear trai n
N:1
9
dri ve
motor
T
s
Z = ? Z =
1
2
9
2
1
9
1
9
2
9 = -10 9
2 1
wi th the rotati onal di recti ons as shown:
1
sJ
1
Impedance Relationships across Two-Port Elements
(1) The Transformer:
Let an element (or system) with impedance Z
1
be connected across a transformer as
shown, then the impedance seen from side 1 of the transformer is
V
1
(s)
Z = .
F
1
(s)
At node (b)
f
2
= f
Z
1
v
2
= v
Z
1
so that
Z =
V
1
(s)
=
kV
2
(s)
= k
2
V
Z
1
(s)
= k
2
Z
1
F
1
(s) (1/k)F
2
(s) F
Z
1
(s)
Example 2
Find the apparent inertia of an inertia J with a step-gear box with a 10:1
ratio.
With the above denition k =
1
/
2
= 0.1, so that
0.01 1
Z
in
= k
2
Z
1
(s) = =
Js (100J)s
or in other words the inertia J is seen at the input shaft as an equivalent
inertia 100J.
168
V
s
4
3
Z
1
Z
2
1
2
output i s v
Z
2
(2) The Gyrator:
Z = ? Z
1
2
= >
v = kf
1
2
f = - v
1
2
1
k
1
At node (b)
f
2
= f
Z
1
v
2
= v
Z
1
so that
Z =
V
1
(s)
=
kF
2
(s)
= k
2
F
Z
1
(s)
= k
2
1
= k
2
Y
1
.
F
1
(s) (1/k)V
2
(s) v
Z
1
(s) Z
1
The nature of the apparent impedance as seen at the input has been changed to its
reciprocal. The result is that an A-type element appears to be a T-type element when
connected behind a gyrator, and vice-versa.
No example is given here because there are no naturally occurring gyrators
in the energy domains covered this term.
Impedance Based Modeling with Two-Port Elements
Consider a Thevenin source coupled to a load Z
2
through a transformer.
The transformer provides the constraints
v
4
= kv
3
f
4
= (1/k)f
3
169
V
s
Z
1
k Z
2
2
(a) If Z
2
is reected to the l.h. side of the transformer:
k
2
Z
2
V
4
(s) = V
s
(s)
Z
1
+ k
2
Z
2
but
1
V
Z
2
= v
3
= v
4
k
so that the transfer function H(s) is
V
Z
2
kZ
2
H(s) = = .
V
s
(s) Z
1
+ k
2
Z
2
(b) Alternatively, we can transfer all elements to the r.h. side of the transformer
1
v
Z
1
= v
3
= v
4
k
From loop (1)
v
Z
1
+ v
4
V
s
= 0
and
1
v
Z
2
= (v
3
v
Z
1
) but v
Z
1
= Z
1
f
Z
1
k
1
= (V
s
Z
1
f
4
) (f
Z
1
= f
4
)
k
1
Z
1

=
k
V
s

kZ
2
V
Z
1
so that
1 Z
1

1
V
Z
2
1 + = V
s
k Z
2
k
or
V
Z
2
kZ
2
H(s) = = .
V
s
(s) Z
1
+ k
2
Z
2
which is the same result as above.
Example 3
Use this result to nd the transfer function
J
(s)
H(s) =
V
in
(s)
for the system shown below, which includes a dc servo motor, driven from a
voltage source V
in
(s), driving an inertial load J and bearing damping B
1610
R
L
J
9
Motor
+
-
V (t)
i
s
Beari ng B
Turntabl e
R
B
J V (t)
s
L
1
2
v = 0
9 = 0
Combine the series and parallel impedances and redraw the graph
V
s
Z
1
Z
2
Z = Ls + R
1
Z =
2
1
Js + B
where Z
1
= R + sL, and Z
2
= 1/(Js + B). Assume that for the motor v
b
= v
1
=
K
m
m
. From the previous result
J
(s) kZ
2
H(s) = =
V
in
(s) Z
1
+ k
2
Z
2
K
m
/(Js + B)
=
(R + sL) + K
2
/(Js + B)
m
K
m
H(s) =
JLs
2
+ (RJ + BL)s + (BR + K
2
)
m
1611
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
a s + a s + . . . + a s + a
n
n
n-1
n-1 1 0
b s + b s + . . . + b s + b
m
m
m-1
m-1 1 0
Us Ys
a s + a s + . . . + a s + a
n
n
n-1
n-1 1 0
b s + b s + . . . + b s + b
m
m
m-1
m-1 1 0
Us Ys 1 Xs
a s + a s + . . . + a s + a
n
n
n-1
n-1 1 0
b s + b s + . . . + b s + b
m
m
m-1
m-1 1 0
Us Ys
1
Fs

Spring Term 2008
Lecture 17
1
Reading:
Nise: Chapter 4
1 System Response
Our systems are
Linear and Time-Invariant (LTI), that is they are linear, and their properties do not
change with time, and
are Single-Input Single-Output (SISO).
and are usually represented by a transfer function
which is equivalent to an ordinary dierential equation (ODE) with constant coecients.
dy
n
dy
n1
dy du
m
du
m1
du
a
n
dt
n
+ a
n1
dt
n1
+ . . . + a
1
dt
+ a
0
y = b
m
dt
m
+ b
m1
dt
m1
+ . . . + b
1
dt
+ b
0
u
where the coecients are determined by the system components.
Block diagram algebra allows us to redraw the block diagram as
or alternatively
Consider rst the action of the block containing the numerator polynomial
1
171
a s + a s + . . . + a s + a
n
n
n-1
n-1 1 0
Us 1
Ys

b s + b s + . . . + b s + b
m
m
m-1
m-1 1 0
Ps
Qs
then
dp
m
dp
m1
dp
q(t) = b
m
dt
m
+ b
m1
dt
m1
+ . . . + b
1
dt
+ b
0
p,
which is not a dierential equation. It simply shows that the output of this block is a
weighted sum of the input and its derivatives.
Initially therefore, we concentrate on solving the sub-system represented by the block
or as a dierential equation
dy
n
dy
n1
dy
a
n
dt
n
+ a
n1
dt
n1
+ . . . + a
1
dt
+ a
0
y = u.
From 18.03 we know that the solution to an ODE of this form may be considered to have
two components:
y(t) = y
h
(t) + y
p
(t)
where
y
h
(t) is the solution to the homogeneous equation, that is the solution when u(t) 0,
and
y
p
(t) is a particular solution that satises the ODE for the given u(t).
a) The Homogeneous Solution y
h
(t): Assume that u(t) 0 and that the system
has initial conditions y(0) = C
0
, y(0) = C
1
, y(0) = C
2
, . . . y
(n)
(0) = C
n
, then conjecture a
solution of the form
y
h
(t) = K e
t
so that
d
m
y
(m)
(t) =
y
h
= K
m
e
t
.
dt
m
Substituting into the homogeneous dierential equation gives
n
K
a
n
+ a
n1
n1
+ . . . + a
1
+ a
0

e
t
= 0
Because K = 0, and e
t
= 0 for nite t, we therefore require
n
a
n
+ a
n1
n1
+ . . . + a
1
+ a
0
= 0,
which is the characteristic equation of the system.
172
U(s)
Y(s)
H(s)
as
U(s)
Y(s)
N(s)
D(s)
numerator pol ynomi al
denomi nator pol ynomi al
u(t) = 0
1
y(t)
s + 5s + 6
2
The assumed form of the solution
y
h
(t) = K e
t
is a solution for any value of satisfying the chatacteristic equation. Since the order of the
characteristic polynomial is n, there will be n such roots, and the most general form of y
h
(t)
will be
n
i
t
y
h
(t) =

K
i
e
i=1
where the constants K
i
are determined from the initial conditions, and the
i
are known
as the eigenvalues (characteristic values) of the system. We note that if we write the the
transfer function
N(s)
H(s) =
D(s)
then the characteristic equation (in terms of s is
D(s) = 0.
Example 1
Find the homogeneous response of the system
when y(0) = 2, and y(0) = 1.
The characteristic equation is
s
2
+ 5s + 6 = 0
(s + 3)(s + 2) = 0
so that
s
1
, s
2
= 3, 2
173
0 0. 5 1 1 . 5 2
-6
-4
-2
0
2
4
6
8
7e
-2t
-5e
-3t
total response
t
response
and the response is
y
h
(t) = K
1
e
3t
+ K
2
e
2t
.
K
1
and K2 are found from the initial conditions. When t = 0:
y(0) = 2 = K
1
+ K
2
y(0) = 1 = 3K
1
2K
2
from which K
1
= 5, and K
2
= 7. The complete solution is
y
h
(t) = 5 e
3t
+ 7 e
2t
.
Notice that this system is stable, that is all exponential components in the ho
mogeneous response decay to zero as time t increases.
a) The Particular Solution y
p
(t): The particular solution is often found using the
method of undetermined coecients, in which a solution of a given form, with unknown
coecients, is assumed for the given input function u(t), and the coecients are found to
satisfy the dierential equation.
The following table summarizes the assumed solution forms for some common input
functions:
174
Term in u(t) Assumed form for y
p
(t) Test value
k
kt
n
(n = 1, 2, 3 . . .)
ke
t
ke
jt
k cos(t)
k sin(t)
K
1
K
n
t
n
+ K
n1
t
n1
+ + K
1
t + K
0
K
1
e
t
K
1
e
jt
K
1
cos(t) + K
2
sin(t)
K
1
cos(t) + K
2
sin(t)
0
0

j
j
j
Example 2
Find the particular solution, and the complete solution for the system
1
H(s) =
s
2
+ 5s + 6
when the input u(t) = 4 (for t > 0) and y(0) = 0 and y(0) = 0.
The dierential equation is
d
2
y dy
+ 5 + 6y = u(t).
dt
2
dt
From the above table, for a constant input we assume y
p
(t) = K and substituting
into the dierential equation
d
2
y
p
dy
p
+ 5 + 6y
p
= K
dt
2
dt
0 + 5 0 + 6K = 4
or y
p
(t) = K = 2/3.
From the previous example, y
h
(t) = K
1
e
3t
+ K
2
e
2t
, so the complete solution
is
2
y(t) = y
h
(t) + y
p
(t) = K
1
e
3t
+ K
2
e
2t
+ .
3
With the stated initial conditions (y(0) = 0 and y(0) = 0),
y(0) = 0 = K
1
+ K
2
+ 2/3,
y(0) = 0 = 3K
1
2K
2
from which K
1
= 4/3, and K
2
= 2. The complete solution is therefore
4 2
y(t) = e
3t
2 e
2t
+ .
3 3
175
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
0 T T T T
1 2 3 4
a) Uni t pul ses of di fferent extents b) The i mpul se functi on
1 /T
1
1 /T
2
1 /T
3
1 /T
4
t
@(t)
0
t
@ (t)
T

Spring Term 2008
Lecture 18
1
Reading:
Nise: Sec. 1.5
1 Common Inputs Used in Control System Design and Analysis
Two classes of inputs commonly used to characterize the performance of feedback control
systems are:
(a) The Singularity Functions: These functions have a discontinuity at time t = 0:
(i) The Dirac Delta (Impulse) Function: The delta function is used to charac
terize the response of a system to brief, intense inputs. In the gure below, (a)
shows some unit pulses (pulses with unit area so that if the duration of the pulse
is T , its amplitude is 1/T . The Dirac delta function (t) is the limiting case of
such pulses as T 0. Notice that this implies that the amplitude 1/T .
The strict denition of (t) is
(t) = 0, for t = 0
(t) is undened at t = 0 (lim (t) = )
t 0
0
+
(t)dt = 1
0

(ii) The Unit-Step Function u
s
(t): The unit-step (or Heaviside) function is used
to characterize a systems transient response to a sudden change.
1
181
H(J)
1 . 0
0
0 1 . 0
ti me
J
i ntegrati on
di fferenti ati on
0
ti me
1 . 0
K (J)
J
I
0
ti me
0
@(J)
J
H(J)
1 . 0
0
0 1 . 0
ti me
J
i ntegrati on
di fferenti ati on
0
ti me
1 . 0
K (J)
J
I
0
The denition is
u
s
(t) = 0 for t < 0
= 1 for t > 0
(iii) The Unit-Ramp Function r(t): The unit-ramp function is used to charac
terize a systems ability to follow a time-varying input, and the transient behavior
around a discontinuity in the slope of an input function.
The denition is
r(t) = 0 for t < 0
= t for t > 0
The singularity functions a related to each other by dierentiation and integration, as
shown below:
(b) Sinusoidal Inputs: The response of linear systems to sinusoidal inputs of the form
u(t) = A sin(t + )
182
s + 5s + 6
2
Us
1 Ys
t
u(t) = A si n(Mt)
peri od
0
2F
M
T =
A
is of fundamental importance to control engineering and system dynamics ant will be
studied extensively throughout the course.
Example 1
In practice a machine, described a second-order transfer function G(s), will be
subjected to inputs that change suddenly. Use the unit-step response to deter
mine how long it will take the machines response to settle to a new steady-state
value after a change.
The systems block diagram is
and the governing dierential equation is
d
2
y dy
+ 5 + 6y = u(t).
dt
2
dt
The step response assumes that the system is at rest at time t = 0, that is
y(0) = 0 and y(0) = 0, and that the input u(t) = u
s
(t).
The solution is
y(t) = y
h
(t) + y
p
(t)
where y
h
(t) is the homogeneous solution, and y
p
(t) is the particular integral. The
characteristic equation is
2
+ 5 + 6 = ( + 3)( + 2) = 0
and
y
h
(t) = C
1
e
3t
+ C
2
e
2t
.
Assume y
p
(t) = K and substitute into the dierential equation
0 + 0 + 6K = 1
183
0 1 2 3
0
0. 05
0. 1
0. 1 5
0. 2
R
e
s
p
o
n
s
e
t (sec)
y = 1/6
ss
i ni ti al sl ope i s zero
Js + 1
1
y(t)
u(t) = A si n(Mt)
or y
p
(t) = 1/6. The complete solution is
y(t) = C
1
e
3t
+ C
2
e
2t
+ 1/6.
At time t = 0
y(0) = C
1
+ C
2
+ 1/6 = 0
y(0) = 3C
1
2C
2
+ 0 = 0
giving C
1
= 1/3 and C
2
= 1/2, so that the systems response to a unit-step
input is
1 1
y
step
(t) =
3
e
3t

2
e
2t
+ 1/6.
The response is shown below, and indicates that it takes this system 2.53 seconds
to respond to the step.
Example 2
Find the steady-state response of a rst-order system to a sinusoidal input u(t) =
A sin(t).
The dierential equation is
dy
+ y = u(t)
dt
and assume the complete solution is y(t) = y
h
(t) + y
p
(t) as in the previous
example. The characteristic equation is
+ 1 = 0
184
B
MJ
1
1 + MJ
2
1
M
J
2 si n(B) =
MJ
cos(B) =
1 + MJ
2
1
from which
y
h
(t) = Ce
t/
and assume
y
p
(t) = K
1
cos(t) + K
2
sin(t)
In steady-state we assume that y
h
(t) = C
1
e
t/
has decayed to zero, and
y
ss
(t) = y
p
(t) = K
1
cos(t) + K
2
sin(t)
Substitution of y
p
(t) into the dierential equation gives
(K
1
sin(t) + K
2
cos(t)) + (K
1
cos(t) + K
2
sin(t)) = A sin(t)
or
(K
2
+ K
1
) cos(t) + (K
1
+ K
2
) sin(t)
and comparing coecients
K
2
+ K
1
= 0
K
1
+ K
2
= A,
or
A A
K
1
= , K
2
=
1 + ()
2
1 + ( )
2
so that
A
y
ss
(t) = y
p
(t) = (sin(t) cos(t))
1 + ()
2
A

1

=
1 + ()
2
1 + ()
2
sin(t)
1 + ()
2
cos(t)
A
=
1 + ()
2
(cos() sin(t) sin() cos(t)) .
using the following triangle
Then the system sinusoidal response is
A
y
ss
(t) =
1 + ()
2
sin(t )
where = tan
1
( ). We note the following
185
(a) The steady-state sinusoidal response is a sinusoid of the same frequency as
the input.
(b) There is a phase shift (lag) between the input and output = tan
1
().
(c) The amplitude of the output is a function of the input frequency .
(d) At low frequencies ( 0), the response is y
ss
(t) A sin(t) and the
amplitude approaches that of the input.
At very high frequencies ( ), the response is y
ss
(t) (A/) sin(t
/2) and the response amplitude becomes very small.
When = 1/, y
ss
(t) = (A/
2) sin(t /4), that is the amplitude is

reduced by a factor of 0.707, and the phase shift is 45
.
186
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008

Spring Term 2008
Lecture 19
1
Reading:
Nise: Chapter 4.
1 System Poles and Zeros
Consider a system with transfer function
N(s)
H(s) = .
D(s)
If we factor the numerator and denominator polynomials and write
H(s) = K
(s z
1
)(s z
2
) . . . (s z
m
)
(s p
1
)(s p
2
) . . . (s p
n
)
where
p
1
, p
2
, . . . , p
n
are the roots of the characteristic polynomial D(s), and are known as
the system poles,
z
1
, z
2
, . . . , x
m
are the roots of the numerator polynomial N(s), and are known as the
system zeros.
Note that because the coecients of N(s) and D(s) are real (they come from the modeling
parameters), the system poles and zeros must be either
(a) purely real, or
(b) appear as complex conjugates
and in general we write
p
i
, or z
i
=
i
+ j
i
.
Example 1
Find the poles and zeros of the system
5s
2
+ 10s
G(s) =
s
3
+ 5s
2
+ 11s + 5
5s(s + 2)
=
(s + 3)(s
2
+ 2s + 5)
5s(s + 2)
=
(s + 3)(s + (1 + j2))(s + (1 j2))
1
191
s-pl ane
{s}
{s}
jw
s
o o
x
x
x
-3 -2 -1
-j2
j2
s = s + jw
so that we have
(a) a pair of real zeros at s = 0, 2 and
(b) three poles at s = 3, 1 + j2, and s = 1 j2.
The system poles and zeros completely characterize the transfer function (and there
fore the system itself) except for an overall gain constant K:
m
(s z
i
)
G(s) = K
i=1
n
i=1
(s p
i
)
1.1 The Pole-Zero Plot
The values of a systems poles and zeros are often shown graphically on the complex s-plane
in a pole-zero plot. For example, the poles and zeros of the previous example are drawn:
where the pole positions are denoted by an x, and the zeros are drawn as an o. The gure
shows zeros at s = 0, 2, and poles at s = 3, 1 j2.
The pole-zero plot is used extensively throughout control theory and system dynamics
to provide a qualitative indication of the dynamic behavior of systems.
Aside: In MATLAB a system may be specied by its poles and zeros using the
function zpk(zeros, poles, gain), for example
sys = zpk([0 2], [-3, -1+i*2, -1-i*2], 5)
step(sys)
will plot the step response of the system in the previous example.
192
s-pl ane
jM
I x x
-3 -4
s = I + jM
generates a component e
-4t
generates a component e
-3t
The characteristic equation of a system is
D(s) = (s p
1
)(s p
2
) . . . (s p
n
) = 0
so that the poles are the system eigenvalues, and the form of the homogeneous response is
dictated by the poles:
n
p
i
t
y
h
(t) =

C
i
e
i=1
(when the poles are distinct), and the constants C
i
are determined by the initial condi
tions.
Example 2
Find the poles and hence the homogeneous response components of the system
12
G(s) =
s
2
+ 7s + 12
The characteristic equation is
D(s) = (s + 4)(s + 4) = 0
and the poles are s = 3, 4
The homogeneous response components are therefore y
1
(t) = C
1
e
3t
and y
2
(t) =
C
2
e
4t
, where C
1
and C
2
are dened by the initial conditions.
Note: The poles do not specify the amplitude of the modal components in the
response. They simply indicate the nature of the response components.
193
s-pl ane
jM
I
o
o
x
x
x
-4
-j4
-1
-j2
j2
s = I + jM
j4
x
1.2 Complex Poles and Zeros
We have noted that in general s = + j, and that poles and zeros may appear as complex
conjugate pairs:
p
i,i+1
=
i
j
i
z
k,k+1
=
k
j
k
For example the pole-zero plot
corresponds to the transfer function
G(s) = K
(s j4)(s + j4)
s(s + 4)(s + (1 + j2))s + (1 j2))
s
2
+ 16
= K
s(s + 4)(s
2
+ 2s + 5)
s
2
+ 16
= K
s
4
+ 16s
3
+ 13s
2
+ 2s
The homogeneous response we will have a pair of complex exponential terms associated with
each pair of conjugate pair of poles, such as
. . . + C
i
e
(
i
+j
i
)t
+ C
i+1
e
(
i
j
i
)t
. . .
but C
i
and C
i+1
are also complex (say a jb), so we can write
C
i
e
(
i
+j
i
)t
+ C
i+1
e
(
i
j
i
)t
= (a + jb)e
i
t
e
j
i
t
+ (a jb)e
i
t
e
j
i
t
= ae
i
t
e
j
i
t
+ e
j
i
t

jbe
i
t
e
j
i
t
e
j
i
t

Eulers formulas state
jt
+ e
jt
) cos(t) =
2
1
(e

e
jt
= cos(t) + j sin(t)
sin(t) =
1
(e
jt
e
jt
)
or
e
jt
= cos(t) j sin(t)
2j
194
O(J)
J
)A
)A si n(MJ)
IJ
IJ
s-pl ane
jM
I o o
x
x
x
-2 -1
-j2
j2
s = I + jM
so that the contribution of the complex conjugate pole pair to the homogeneous response
may be written
y
i,i+1
(t) = 2ae
i
t
cos(
i
t) + 2ae
i
t
sin(
i
t))
= 2
a
2
+ b
2
e
i
t

a
a
cos(
i
t) +
a
b
sin(
i
t)

2
+ b
2 2
+ b
2
= A
i
e
i
t
sin(
i
t +
i
)
where
A
i
= 2
a
2
+ b
2
and
i
= tan
1
a

.
b
which is shown below (for
i
< 0).
Example 3
Find and plot the poles and zeros of
7s + 14
G(s) =
s
3
+ 3s
2
+ 7s + 5
and then determine the modal response components of this system.
s + 2 s + 2
G(s) = 7 = 7
(s + 1)(s
2
+ 2s + 5 (s + 1)(s + (1 + j2))(s + (1 j2))
The pole-zero plot is
195
0
unstabl e regi on stabl e regi on
I
jM
and the modal components are (1) Ce
t
(corresponding to the pole at s = 1),
and (2) Ae
t
sin(2t + ) (corresponding to the complex conjugate pole pair at
s = 1 j2), and where the constants C, A, and are determined from the
initial conditions.
Note: A pair of purely imaginary poles (on the imaginary axis of the s-plane)
implies = 0 and there will be no decay. The system will act as a pure oscillator.
The eect of pole locations in the s-plane on the modal response components is summa
rized in the gure below:
We note
(a) Poles in the left-half of the s-plane (the lhp), that is < 0, generate components that
decay with time.
(b) Conversely, poles in the right-half s-plane (the rhp), that is > 0 generate components
that grow with time.
(c) Poles that lie on the imaginary axis ( = 0) generate components that are purely
oscillatory, and neither grow nor decay with time.
(d) A pole at the origin of the s-plane (s = 0+j0), generates a component that is a constant.
In addition we note that oscillatory frequency and decay rate is determined by the distance
of the pole(s) from the origin.
(e) The rate of decay/growth is determined by the real part of the pole = {s}, and
poles deep in the lhp generate rapidly decaying components.
196
I
jM
i ncreasi ng decay rate
i ncreasi ng
frequency
i ncreasi ng
frequency
(f) For complex conjugate pole pairs, the oscillatory frequency is determined by the imagi
nary part of the pole pair = {s}.
1.3 System Stability
A system is dened to be unstable if its response from any nite initial conditions increases
without bound. Since
n
p
i
t
Y
h
(t) =

C
i
e
i=1
a system will be unstable if any component of y
h
(t) increases without bound, leading to the
following statements:
(a) A system is unstable if any pole has a positive real part (ie lies in the rhp), or
equivalently
(b) For a system to be stable, all poles must lie in the lhp.
A system with poles on the imaginary axis (with no poles in the rhp) is dened to be
marginally stable since its homogeneous response from arbitrary initial conditions will neither
decay to zero or increase to innity.
197
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
s-pl ane
jM
x
a si ngl e real pol e
1
J
-
I

Spring Term 2008
Lecture 20
1
Reading:
Nise: Secs. 4.1 4.6 (pp. 153 - 177)
1 Standard Forms for First- and Second-Order Systems
These are (a) all pole system (with no zeros), and (b) have unity gain (lim
t
y
step
(t) = 1).
1.1 First-Order System:
We dene the rst-order standard form as
1
G(s) = ,
s + 1
where the single parameter is the time constant. As a dierential equation
dy
+ y = u(t).
dt
and the system has a single real pole at s = 1/.
1
201
0
T 4t
0
1
t
y (t)
step
0. 98
t
1
t
0. 1
0. 9
R
T s
i ni ti al sl ope i s
The step response is
y
step
= L
1
1 1/

= 1 e
t/
s s + 1/
1.1.1 Common Step Response Descriptors:
(a) Settling Time: The time taken for the response to reach 98% of its nal value. Since
y
step
(t) = 1 et/
and e
4
= 0.0183 0.02, we take
T
s
= 4
as the denition of T
s
.
(b) Rise Time: Commonly taken as the time taken for the step response to rise from
10% to 90% of the steady-state response to a step input. It is found from the step
response as follows
0.1 = 1 e
t
0.1
/
t
0.1
= ln(0.9)
0.9 = 1 e
t
0.9
/
t
0.9
= ln(0.1)
T
R
= t
0.9
t
0.1
= (ln(0.1) ln(0.9)) = 2.2
1.2 Second-Order Systems
The standard unity gain second-order system has a transfer function
2
G(s) =
n
s
2
+ 2
n
s +
2
n
with two parameters (1)
n
the undamped natural frequency, and (ii) the damping
ratio ( 0). The system poles are the roots of s
2
+ 2
n
s +
n
2
= 0, that is
p
1
, p
2
=
n

n
2
1,
leading to four cases
202
s-pl ane
jM
I
x
x
x
x
x
xx
real pol e
coi nci dent pol es
conj ugate pol es
i magi nary pol es
jw
s
s-pl ane
x
x
jw
-jw
n
n
-w
n
-zw
n
w
n
w
n
jw
n 1 - z
2
q
w
n 1 - z
2
zw
n
w
n
-1
q = cos z
i) > 1 the poles are real and distinct
p
1
, p
2
=
n

n
2
1,
ii) = 1 the poles are real and coincident
p
1
, p
2
=
n
,
iii) 0 < < 1 the poles are complex conjugates
p
1
, p
2
=
n
j
n
1
2
, or
(iv) = 0 the poles are purely imaginary
p
1
, p
2
= j
n
.
1.2.1 Pole Positions For an Underdamped Second-Order System
p
1
, p
2
=
n
j
n
1
2
and when plotted on the s-plane
we note that
203
jw
s
s-pl ane
w i ncreasi ng
n
l i nes of constant w
n
jw
s
s-pl ane
l i nes of constant z
z = 0
z = 0
z = 1
z = 1
J
O (J)
IJAF
no overshoot
z > 1
(a) The poles lie at a distance
n
from the origin, and
(b) The poles lie on radial lines at an angle
= cos
1
()
as shown above.
The inuence of and
n
on the pole locations may therefore be summarized:
1.2.2 Step Responses
(a) The over damped case ( > 1)
y
step
(t) = 1 C
1
e
p
1
t
C
2
e
p
2
t
where the constants C
1
and C
2
are determined from p
1
and p
2
.
(b) The critically damped case ( = 1) With two coincident poles p
1
= p
2
= p, the
step response takes a special form
y
step
(t) = 1 C
1
e
pt
C
2
te
pt
204
J
O (J)
IJAF
z < 1
damped osci l l atory response
overshoot
J
O (J)
IJAF
no overshoot
z = 1
(c) The under damped case (0 < < 1) With a pair of complex conjugate poles
p
1
, p
2
=
n
j
n
1
2
the step response becomes oscillatory
e
n
t
y
step
(t) = 1
1
2
cos
1
2
t

where

= tan
1
1
2
and if we dene the damped natural frequency
d
as
d
=
n
1
2
we can write the step response as
e
n
t
y
step
(t) = 1 (cos (
d
t ))
1
2
205
2
1
0
J
O (J)
IJAF
z = 0
pure osci l l atory response
t
y (t)
step
t
1
0
no overshoot
z > 1
0. 9
0. 1
t
0. 1
t
0. 9
T
R
1
0
0. 9
0. 1
t
0. 1
t
0. 9
T
R
z < 1
y (t)
step
(d) The undamped case ( = 0) In this case
n
G(s) =
s
2
+
2
n
and the poles are p
1
, p
2
= j
n
. Then
y
step
(t) = 1 cos (
n
t)
Note: For any second-order system, the initial slope of the step response is zero,
since by denition the system is at rest at time t = 0, that is y
step
(0) = 0, and
y
step
(0) = 0.
1.2.3 Step Response Based Second-Order System Specications
(a) Rise Time (T
R
): Applies to over- and under-damped systems. As in the case of
rst-order systems, the usual denition is the time taken for the step response to rise
from 10% to 90% of the nal value:
For a second-order system there is no simple (general) expression for T
R
. The following
gure - from Nise, Fig. 4.16, (p. 172) is derived empirically:
206
t
y (t)
step
1
0
T
P
z < 1
y
peak
0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9
1
1 . 2
1 . 4
1 . 6
1 . 8
2
2. 2
2. 4
2. 6
2. 8
dampi ng rati o
z
R
i
s
e
t
i
m
e
n
o
r
m
a
l
i
z
e
d
f
r
e
q
u
e
n
c
y
(b) Peak Time (T
p
): Applies only to under-damped systems, and is dened as the time
to reach the rst peak of the oscillatory step response.
T
p
is found by dierentiating the step response y
step
(t), and equating to zero. (See Nise
p. 170 for details.)

T
p
= =
1
2
d
** Transient response specications continued in Lecture 21. **
207
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
t
y = 1
y (t)
step
z < 1
0
y (t) = 1 -
step
e
1 - z
2
-zw t
n
cos(w t - f)
d
y
peak
ss
T
p
overshoot

Spring Term 2008
Lecture 21
1
Reading:
Nise: Secs. 4.6 4.8 (pp. 168 - 186)
1 Second-Order System Response Characteristics (contd.)
1.1 Percent Overshoot
The height of the rst peak of the response, expressed as a percentage of the steady-state
response.
%OS =
y
peak
y
ss
100
y
ss
At the time of the peak y(T
p
)
y
peak
= y(T
p
) = 1 + e
(/
1
2
)
and since y
ss
= 1
%OS = e
(/
1
2
)
100.
Note that the percent overshoot depends only on .
Conversely we can nd to give a specic percent overshoot from the above:
ln (%OS/100)
=
2
+ ln
2
(%OS/100)
1
211
Example 1
Find the damping ratio that will generate a 5% overshoot in the step response
of a second-order system.
Using the above formula
ln (%OS/100) ln(0.05)
= = = 0.69
2
+ ln
2
(%OS/100)
2
+ ln
2
(0.05)
Example 2
Find the location of the poles of a second-order system with a damping ratio
= 0.707, and nd the corresponding overshoot.
The complex conjugate poles line on a pair of radial lines at an angle
= cos
1
0.707 = 45

from the negative real axis. The percentage overshoot is
e
(/
1
2
)
%OS = 100
e
(0.707/
10.5)
= 100
= 4.3% ( 5%)
The value = .707 =
2/2 is a commonly used specication for system design

and represents a compromise between overshoot and rise time.
1.2 Settling Time
The most common denition for the settling time T
s
is the time for the step response y
step
(t)
to reach and stay within 2% of the steady-state value y
ss
. A conservative estimate can be
found from the decay envelope, that is by nding the time for the envelope to decay to less
than 2% of its initial value,
e
n
t
1
2
< 0.02
giving
ln(0.02
1
2
)
T
s
=
n
or
4
T
s

n
for
2
1.
212
jM
I
s-pl ane
x
x
45
o
j6. 28
-j6. 28
j8. 88
-j8. 88
8
.
8
8
-6. 28
desi red pol e
posi ti ons
Example 3
Find (i) the pole locations for a system under feedback control that has a peak
time T
p
= 0.5 sec, and a 5% overshoot. Find the settling time T
s
for this system.
From Example 2 we take the desired damping ratio = 0.707. Then

T
p
= = 0.5 s
1
2
so that

n
= = = 8.88rad/s.
T
p
1
2
0.5
1 0.5
The pole locations are shown below:
Then
p
1
, p
2
= 8.88 cos

4

j8.88 sin

4

= 6.28 j6.28
The indicated settling time T
s
from the approximate formula is
4 4
T
s
= = 0.64 s.
n
0.707 8.88
Note that in this case does not meet the criterion
2
1 and the full expression
ln(.02
1
2
) ln(.02
1 0.5)
T
s
=
n
=
0.5 8.88
= 0.68 s
gives a slightly larger value.
213
5
4s+1
5
4s + 2s +5
+
-
yt
ut
2 Higher Order Systems
For systems with three or more poles, the system be analyzed as a parallel combination of
rst- and second-order blocks, where complex conjugate poles are combined into a single
second-order block with real coecients, using partial fractions. The total system output is
then the superposition of the individual blocks.
Example 4
Express the system
5
G(s) =
(s + 1)(s
2
+ 2s + 5)
as a parallel combination of rst- and second-order blocks.
5 A Bs + C
G(s) = = +
(s + 1)(s
2
+ 2s + 5) s + 1 s
2
+ 2s + 5
5 1 5 s + 1
=
4 s + 1

4 s
2
+ 2s + 5
using partial fractions. The system is described by the following block diagram
and the response to an input u(t) may be found as the (signed) sum of the
responses of the two blocks.
3 Some Fundamental Properties of Linear Systems
3.1 The Principle of Superposition
For a linear system at rest at time t = 0, if the response to an input u(t) = f(t) is y
f
(t), and
the response to a second input u(t) = g(t) is y
g
(t), then the response to an input that is a
linear combination of f(t) and g(t), that is
u(t) = af(t) + bg(t)
where a and b are constants is
y(t) = ay
f
(t) + by
g
(t).
214
/(I) B(J) y(J)
i mpl i es /(I)
@B
@J
@O
@J
/(I) B(J) y(J)
i mpl i es /(I)
B@J
O@J
J
i ntegrati on
di fferenti ati on
0
ti me
1 . 0
K J
J
I
0
ti me
0
@J
J
K J
1 . 0
0
0 1 . 0
ti me
J
i ntegrati on
di fferenti ati on
H

3.2 The Derivative Property
f
(t),
then the response to an input that is the derivative of f(t), that is
df
u(t) =
dt
is
dy
f
y(t) = .
dt
3.3 The Integral Property
f
(t),
then the response to an input that is the integral of f(t), that is
t
u(t) = f(t)dt
0
is
t
y(t) = y
f
(t)dt.
0
Example 5
We can use the derivative and integral properties to nd the impulse and ramp
responses from the step response. We have seen
therefore
d
y
(t) = y
step
(t)
dt
t
y
r
(t) = y
step
(t)dt
0
215
KNs ut Ot
,s
1 Nt
KNs Ot
,s
1
Lt
ut
For example, consider
b
G(s) =
s + a
with step response
b
1 e
at

. y
step
=
a
The impulse response is
d
y
(t) = y
step
(t) = be
at
,
dt
and the ramp response is
t

b

1
1 e
at

y
r
(t) = y
step
(t)dt = t +
0
a a
4 The Eect of Zeros on the System Response
Consider a system with a transfer function:
G(s) = K
N(s)
= K
s
m
+ b
m1
s
m1
+ + b
1
s + b
0
.
D(s) s
n
+ a
n1
s
n1
+ + a
1
s + a
0

N(s), which denes the system zeros, is associated with the RHS of the dierential equation,
while D(s) is derived from the LHS of the dierential equation. Therefore N(s) does not
aect the homogeneous response of the system.
We can draw G(s) as cascaded blocks in two forms:
or
In this case we consider the all-pole In this case we consider the all-pole sys
system 1/D(s) to be excited by x(t), tem 1/D(s) to be excited by u(t) di
which is a superposition of the deriva- rectly to generate x(t), and the out
tives of u(t) put is formed as a superposition of the
derivatives of v(t)
m
x(t) = K

b
k
d
k
u
m
k=0
dt
k
y(t) = K

b
k
d
k
v
dt
k
k=0
216
s-pl ane
I
x
x
o
-10
-1
j2
-j2
KNs ut Ot
,s
1 Nt
Example 6
Find the step response of
s + 10
G(s) = .
s
2
+ 2s + 5
Splitting up the transfer function
1
N(s) = s + 10, D(s) =
s
2
+ 2s + 5
and
n
=
5, and = 1/
5.
Method 1: For the case,
if u(t) = u
s
(t), the unit-step function
du
x(t) = + 10u = (t) + 10u
s
(t)
dt
and for the all-pole system 1/D(s)
1
1

y
step
(t) = 1 e
t
cos(2t) e
t
sin(2t)
5 2
1
y
(t) = e
t
sin(2t).
2
For the complete system
y(t) = y
(t) + 10y
step
(t)
1
= 2 2e
t
cos(2t) e
t
sin(2t)
2
Method 2: For the case,
217
KNs Ot
,s
1
Lt
ut
from above the step-response to the all-pole system 1/D(s) is
1
1

v(t) = 1 e
t
cos(2t) e
t
sin(2t)
5 2
and the system output is
dv
y(t) = + 10v
dt
1 10
1

= e
t
sin(2t) + 1 e
t
cos(2t) sin(2t)
2 5 2
1
= 2 2e
t
cos(2t) e
t
sin(2t)
2
which is the same as found in Method 1.
218
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
K
s + 2s 5
s + b
U(s)
V(s) Y(s)
u(t)
v(t)
y(t)
I
jM
x
x
o
b i ncreasi ng
-b

Spring Term 2008
Lecture 22
1
Reading:
Nise: 4.1 4.8
1 The Time-Domain Response of Systems with Finite Zeros
Consider a system:
K(s + b)
Gs = ,
s
2
+ 2s + 5
we have seen that we can consider this as two cascade blocks
Then if the response of the a system 1/D(s) is v(t), then
dv
y(t) = + bv(t)
dt
and as the zero (at s = b) moves deeper into the l.h. s-plane,, the relative contribution of
the derivative term decreases
and the system response tends toward a scaled version of the all pole response v(t).
In general, the presence of the derivative terms in the response means that:
1
221
0 1 2 3 4 5 6 7
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
z = 1
z = 2
z = 3
allpole
The response is faster (shorter peak-time T
P
and rise-time T
R
).
Greater overshoot in the response (if any). A zero may cause overshoot in the response
of an over-damped second-order system.
Example 1
The following MATLAB step response compares the response for the under-
damped system
5
G(s) =
s
2
+ 2s + 5
with similar unity-gain systems with zeros at s = 1, 2, 3:
5(s + 1) 5/2(s + 2) 5/3(s + 3)
G(s) = , G(s) = , G(s) =
s
2
+ 2s + 5 s
2
+ 2s + 5 s
2
+ 2s + 5
Note the increase in the overshoot, and the decrease in T
P
as the zero approaches
the origin.
222
0 0.5 1 1.5 2 2.5
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
12(s + 1)
s + 7s + 12
s + 7s + 12
12
2
2
Example 2
The following MATLAB step response compares the response for the unity-gain
overdamped system
12
G(s) =
s
2
+ 7s + 12
with two real poles at s = 3 and s = 4 with the similar system with a zeros
at s = 1:
12(s + 1)
G(s) =
s
2
+ 7s + 12
Note the overshoot caused by the zero, but that the overshoot is not oscillatory.
Clearly the rise-time T
R
is much shorter for the system with the zero.
2 The Time-Domain Response of Systems where the Order of
the Numerator equals the Order of the Denominator
Consider systems of the form
b
n
s
n
+ b
n1
s
n1
+ . . . + b
1
s + b
0
G(s) =
a
n
s
n
+ a
n1
s
n1
+ . . . + a
1
s + a
0
223
b - a
s + a
1
+
+
y(t)
u(t)
U(s)
Y(s)
di rect feed-through from i nput to output
where the degree of the numerator equals that of the denominator. In such systems it is
possible to do polynomial division and write the transfer function as
N(s) N
(s)
G(s) = = K +
D(s) D(s)
where N
(s) is a polynomial of degree less than that of D(s).

For example, a system with transfer function
s + a
G(s) =
s + b
may be written
b a
G(s) = 1 + ,
s + a
which may be represented in block-diagram form
showing a direct feed-through of the input into the output. In other words, when the order of
the numerator is the same of the denominator the input will appear directly as a component
of the output.
The step-response y
step
(t) of this system will therefore be
b a
1 e
at
y
step
(t) = u
s
(t) +
a
where u
s
(t) is the unit-step (Heaviside) function.
Note:
That y
step
(0
+
) = 1, that is there is a step transient in the response (which does not
occur if the order of N(s) is less than that of D(s)).
The steady-state step response y
ss
= b/a, and if b > a then y
ss
> 1, while if a > b
y
ss
< 1.
The following MATLAB plot shows the step responses for the two systems
s + 6 s + 2
G(s) = (b > a), and G(s) = (a > b)
s + 4 s + 4
with step responses
y
step
(t) = 1 +
2
1 e
4t

and y
step
(t) = 1
2
1 e
4t

4 4
224
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
0 0.5 1 1.5
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
s + 6
s + 4
s + 2
s + 4
Initial transient at t=0
G(s) =
G(s) =
Find the step response of the following electrical circuit:
+
-
C
R
R
1
2 v
V
o
1mF
10kW
20kW
Example 3
V
o
(s) s + 1/R
1
C
G(s) = =
V (s) s + (R
1
+ R
2
)/R
1
R
2
C
and with the values shown
V
o
(s) s + 100 50
G(s) = = = 1 .
V (s) s + 150 s + 150
The step response is therefore
50
1 e
150t

=
2 1
e
150t
y
step
= 1 +
150 3 3
which is plotted below:
225
0.005 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
0
0.2
0.4
0.6
0.8
1
1.2
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
Example 4
Find the step response of the following third-order system:
2s
3
+ 17s
2
+ 13s + 12
G(s) =
s
3
+ 7s
2
+ 6s + 5
3s
2
+ s + 2
= 2 +
s
3
+ 7s
2
+ 6s + 5
showing a direct feed-through term of amplitude two. From Maple-Syrep, the
step response is
y
step
(t) = 2.40.5307e
6.157t
+0.1307e
0.4213t
cos(0.7966t)0.2667e
0.4213t
sin(0.7966t)
from which y
step
(o
+
) = 2, and y
ss
= 2.4. The step response is plotted below.
226
0 2 4 6 8 10 12 14 16
0
0.5
1
1.5
2
2.5
3
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
227
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
m
K
B
B
1
2
V
v
m
s
I
jM
ox x
-3 -2

Spring Term 2008
Lecture 23
1
Reading:
Nise: 4.7, 4.8
1 Pole/Zero Cancelation:
Consider the following mechanical system:
The transfer function relating v
m
to V
s
is
v
m
(s) B
1
s + K
G(s) = =
V
s
(s) ms
2
+ (B
1
+ B
2
)s + K
which is clearly second-order. However, if the parameter values are m
1
= 1/3 kg, K = 2
N/m, B
1
= 1 N-s/m, B
2
= 2/3 N-s/m the transfer function becomes
s + 2 3(s + 2)
G(s) = =
(1/3)s
2
+ (5/3)s + 2 (s + 2)(s + 3)
with a coincident pole and zero at s = 2.
1
231
s+2
3
s + 5s + 6
x(t)
V (t) = u (t)
v (t)
s s m
Clearly, with these values, cancelation takes place in the transfer function and
3
G(s) =
s + 3
This phenomenon is known as pole/zero cancelation. With these values the response will be
identical to a rst-order system with a pole at s = 3.
Lets compute the step response of this system to understand what has happened. We
start by breaking the system into two cascaded blocks as we have discussed previously:
When the input V
s
(t) = u
s
(t), the unit-step (Heaviside) function, the intermediate variable
x(t) is
du
s
(t)
x(t) = + 2u
s
(t) = (t) + 2u
s
(t)
dt
where (t) is the Dirac delta function. The system output v
m
(t) will therefore be
v
m
(t) = y
(t) + 2y
s
(t)
where y
(t) is the impulse response of the all-pole system to the right, and y
s
(t) is its step
response. You can show for yourself that
y
s
(t) = 0.5 1.5e
2t
+ e
3t
y
(t) = 3e
2t
3e
3t
so that
v
m
(t) = y
(t) + 2y
s
(t) = (3e
2t
3e
3t
) + 2(0.5 1.5e
2t
+ e
3t
)
= 1 e
3t
which is identical to the step response of the reduced system
3
G(s) = .
s + 3
Notice that the modal component e
2t
has been canceled from the response. Although we
have simply demonstrated the cancelation in this example, the following statement is true
for all inputs:
For any system in which one or more poles have been canceled by zeros in the
transfer function, the modal components e
pt
corresponding to the canceled poles
will not appear in the output for any input u(t).
We say that the canceled modes are not excited by the input.
232
2 The Eect of Zeros in the Proximity of a Pole
The previous discussion considered only the complete cancelation of a pole. In many systems
there may be zeros close to a pole but not coincident with it. Lets address this issue by
example again by considering the step response of a second-order system with two real poles
and a single, this time in symbolic form:
ab (s + c)
G(s) =
c (s + a)(s + b)
where the constant ab/c forces this to be a unity-gain system.
The step-response for this system is:
y
s
(t) = 1
b(a c)
e
at
+
a(b c)
e
bt
.
c(a b) c(a b)
Note that the amplitude of each of the modal components is determined by the distance from
the pole to the zero, that is (a c) for the component e
at
, and (b c) for the component
e
bt
, and although this is simply a demonstration The following result is true:
As a zero approaches a pole, the amplitude of the modal component e
pt
correspond
ing to the pole decreases, for any input u(t).
This example also demonstrates the complete elimination of the component e
pt
when the
zero and pole are coincident.
Example 1
Compare the step responses of the system
(12/c)(s + c)
G
1
(s) =
(s + 3)(s + 4)
for
(i) c = 2.95 (ie, a zero close to the pole at s = 3).
(ii) c = 3.50 (ie, a zero mid-way between the two poles).
(iii) c = 2.95 (ie a zero close to the pole at s = 4).
Substitution into the above gives the three step responses as
c = 2.95: y
s
(t) = 1 + 0.068e
3t
1.068e
4t
c = 3.50: y
s
(t) = 1 0.571e
3t
0.429e
4t
c = 4.05: y
s
(t) = 1 1.007e
3t
+ 0.007e
4t
233
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
0.2
0.4
0.6
0.8
1
1.2
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
c = 4.05
c = 3.5
c = 2.95
s+c
(s + 3)(s + 4) c
12
Step response of:
The eect on the amplitude of the component associated with the pole close to
the zero is obvious. The three responses are plotted below:
3 The Response of High-Order Systems
The exact form of the response of third-order and higher systems is not easy to tabulate
and summarize. For example, a third-order system may have three real poles, or two com
plex conjugate poles and a single real pole. The situation becomes even more complicated
for fourth- and higher order systems. Nevertheless the same basic rules apply as we have
developed for rst- and second-order systems:
The total response to an input u(t) will consist of two fundamental components: the
homogeneous response y
h
(t), and a particular solution y
p
(t) that is dened by u(t).
The homogeneous component will consist a set of a superposition of n modal compo
nents of the form e
pt
associated with the system poles, that is
n
p
i
t
y
h
(t) =

C
i
e
i=1
where the constants C
i
are later determined by the assumption that the system is at
rest at t = 0.
234
jM
jM
x x
"sl ow" decay
"rapi d" decay
t
t
The durati on of
each component
depends on i ts
real part
The particular solution is usually determined (if necessary) by the method of undeter
mine coecients, in which it is assumed that y
p
(t) is similar in form to the input. The
value of y
p
(t) is aected by the system zeros.
The total solution is therefore
n
p
i
t
y(t) =

C
i
e + y
p
(t)
i=1
The pole-zero plot essentially tells us what the components of the response will be, but gives
little information about the strength of the components, Never the less it is a very useful
tool in control system design.
4 Model Approximation Dominant Poles:
We have seen that the modal component of a real pole at s = is Ce
t
, and the component
due to a complex conjugate pole pair at s = j is Ce
t
cos(t + ), and that the decay
rate of the component is dened by the real part of the pole. Poles lying deep in the l.h.
plane decay much faster than those close to the imaginary axis.
Poles lying close to the imaginary axis are known as dominant poles because their response
components persist for much longer that those deep within the l.h. plane.
There is often the need to simplify a system model, for example to approximate a third-
order system by a second-order model. This can sometimes be done by retaining only the
dominant pole pair - provided the third pole is suciently well separated from the dominant
pair.
Consider a system shown with transfer function
50 50
G(s) = =
s
3
+ 12s
2
+ 255s + 50 (s + 10)(s
2
+ 2s + 5)
235
jM
jM
N
N
N
-1
j2
-j2
-5
0 1 2 3 4 5 6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e 3rdorder system
2ndorder approximation
which has poles at s = 10, 1 j2, as shown below:
By ignoring the pole at s = 10, and maintaining a unity-gain system we might assert
5
G(s)
s
2
+ 2s + 5
The following gure compares the step response of the two systems.
It can be seen that the responses are similar.
Note: There is a danger in using reduced-order models in closed-loop control system
design. For example the use of a second-order approximation to a real third-order
system will indicate that the system will never become unstable with proportional
control. The physical system, however, will become unstable as the proportional
gain is increased.
For example, the system shown in the above example will be unstable under proportional
control for K
p
> 5, which cannot be predicted from the second-order approximation.
236
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
+
-
R(s)
C(s)
G (s)
G (s)
control l er
pl ant
c
p
+
-
R(s)
C(s) G (s)
G (s)
control l er & pl ant
c p

Spring Term 2008
Lecture 24
1
Reading:
Nise: Chapter 7
1 The Poles and Zeros of Closed-loop systems:
Consider the unity feedback system shown below with a controller G
c
(s) and plant G
p
(s):
Combine the two cascaded blocks to form a single forward transfer function G
f
(s) = G
c
(s)G
p
(s)
and write
N
f
(s)
G
f
(s) =
D
f
(s)
in terms of the numerator polynomial N
f
(s) and denominator polynomial D
f
(s). The closed-
loop transfer function is
G
f
(s) N
f
(s)
G
cl
(s) = =
1 + G
f
(s) D
f
(s) + N
f
(s)
from which we see that
1
241
0 0.2 0.4 0.6 0.8 1 1.2 1.4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
The closed-loop poles are the roots of the characteristic equation N
f
(s) + D
f
(s) = 0.
The closed-loop zeros are the same as the zeros of the forward transfer function.
Example 1
Find the closed-loop transfer function of the plant G
p
(s) = 3/(s + 3) under P-D
control where G
c
= 10 + 2s.
The forward transfer function is
6(5 + s)
G
f
(s) = G
c
(s)G
p
=
s + 3
The closed-loop transfer function is:
N
f
(s) 6(5 + s) 6(s + 5)
6

s + 5
G
cl
(s) = = = =
D
f
(s) + N
f
(s) (s + 3) + 6(5 + s) (7s + 33) 7 s + 33/7
so that the closed -loop pole is at s = 33/7 = 4.7143 and the closed-loop zero
is at s = 5 (the same as the open loop zero dened by the P-D controller).
Aside: The system can be analyzed using the following MATLAB commands:
forward_system = zpk(-5, -3, 6)
closed_loop =feedback(forward_system,1)
pole(closed_loop) %Find the closed-loop system poles
zero(closed_loop) %Find the closed-loop system system zeros
pzmap(closed_loop) %Make a pole-zero plot
step(closed_loop) %Plot the closed-loop step response.
242
The step response is shown below note the initial transient caused by the direct
feed-through.
Now consider a closed-loop system with sensor dynamics H(s)
+
-
R(s)
C(s) G (s)
G (s)
c p
H(s)
sensor
The closed-loop transfer function is
G
f
(s) N
f
(s)D
H
(s)
C
cl
(s) = =
1 + G
f
(s)H(s)) D
f
(s)D
H
(s) + N
f
(s)N
H
(s)
where N
H
(s) and D
H
(s) are the numerator and denominator polynomials of the sensor
transfer function H(s). In this case:
The closed-loop poles are the roots of the characteristic equation
D
f
(s)D
H
(s) + N
f
(s)N
H
(s) = 0.
The closed-loop zeros are the zeros of the forward transfer function, and the poles of
the sensor transfer function.
Example 2
Repeat the previous example with a sensor that has a transfer function H(s) =
10/(s + 10). The forward transfer function is
6(5 + s)
G
f
(s) = G
c
(s)G
p
=
s + 3
and
10
H(s) =
s + 10
The closed-loop transfer function is:
N
f
(s)D
H
(s) 6(5 + s)(s + 10) 6(s
2
+ 15s + 50)
G
cl
(s) = = =
D
f
(s)D
H
(s) + N
f
(s)N
H
(s) (s + 10)(s + 3) + 60(5 + s) (s
2
+ 73s + 330)
so that the closed-loop poles are at s = 4.84, 68.16 and the closed-loop zeros
are at s = 5, 10 (the same as the open loop zero dened by the P-D controller,
and the pole associated with the sensor).
243
0 1 2 3 4 5 6
0
0.2
0.4
0.6
0.8
1
Step Response
Time (sec)
U
n
i
t

s
t
e
p

r
e
s
p
o
n
s
e
desired y = 1
ss
e = 0
e = 0.25
ss
ss
2 Steady-State Errors
In the lab we have considered steady-state errors for both velocity and position control of
the rotary inertia, and we have noted:
There was a nite s.s. error with a constant input under velocity control.
That the s.s. error was eliminated when we used PI (proportional + integral) control.
There was no s.s. error with a constant input for position control.
We will look at the steady state error for two basic inputs
1. The step input. The step response measures the ability of a feedback control system
to regulate the output to a constant input.
The above gure shows the response to a unit step. One response exhibits e
ss
= 0,
while the other shows a constant steady-state error.
2. The ramp input. The steady-state ramp response error is a measure of a feedback
control systems ability to follow a simple time-varying trajectory.
244
+
-
R(s)
C(s) G(s)
E(s)
0 1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
5
6
7
8
9
10
Time (sec)
U
n
i
t

r
a
m
p

r
e
s
p
o
n
s
e
infinite e
zero e
finite e
ss
ss
ss
The above gure shows three responses to a unit ramp input r(t) = t. In one case there
is no steady-state error - as t becomes large, the response follows the input exactly.
In the second case there is a nite steady state error, the response has unit slope but
exhibits a constant oset from the input. The third case shows a response in which
the error is growing without bound, and the steady-state error is innite.
We now look at the whole question of steady-state errors under closed-loop control, and
methods to eliminate them. Consider the unity feedback system:
The error signal E(s) is dened to be E(s) = R(s) C(s), and the transfer function relating
the error to the input command is found from:
C(s) = G)s)E(s)
E(s) = R(s) C(s)
giving
E(s) 1 D(s)
= =
R(s) 1 + G(s) D(s) + N(s)
where N(s) and D(s) are the numerator and denominator polynomials of G(s) respectively.
We recall the nal value theorem:
lim f(t) = lim sF (s)
t s0
245
jM
jM
x
x
x
x o
For zero steady-state error for a
step i nput, there must be at
l east one pol e at the ori gi n.
so that
D(s)

e
ss
= lim e(t) = lim sR(s)
t s0
D(s) + N(s)
Now consider the two cases:
Step input: In this case, when r(t) = u
s
(t), R(s) = 1/s so that

1 D(s)

D(s)

e
ss
= lim s = lim
s 0
s D(s) + N(s)
s 0
D(s) + N(s)

The condition to ensure that e
ss
= 0 therefore must be that
lim D(s) = 0
s 0
If
m
i=1
(s z
i
)
G(s) = K
n
i=1
(s p
i
)
or D(s) =
n
(s p
i
), to ensure that lim
s 0
D(s) = 0 we require that at least one
i=1

of the p
i
= 0 (one or more poles of the system be at the origin). This is equivalent to
saying that the forward transfer function must be of the form
m
i=1
(s z
i
)
G(s) = K
k
n
k 1
s
i=k+1
(s p
i
)
Ramp input: In this case, when the input is a unit ramp r(t) = t, R(s) = 1/s
2

1 D(s)

1 D(s)

e
ss
= lim s
2
= lim
s 0
s D(s) + N(s)
s 0
s D(s) + N(s)

The condition to ensure that e
ss
= 0 therefore must be that
D(s)
lim = 0
s 0
s
246
jM
jM
x
x
x
x o
For zero steady-state error for a
ramp i nput, there must be at
l east two pol es at the ori gi n.
x
If as before
m
i=1
(s z
i
)
G(s) = K
n
i=1
(s p
i
)
or D(s) =
n
(s p
i
), to ensure that lim
s 0
D(s)/s = 0 we require that at least two
i=1

of the p
i
= 0 (one or more poles of the system be at the origin). This is equivalent to
saying that the forward transfer function must be of the form
m
i=1
(s z
i
)
G(s) = K
s
k
i
n
=k+1
(s p
i
)
k 2.
The above argument can be extended as follows:
For zero steady-state error to a waveform with a Laplace transform 1/s
k
,
the forward transfer function must have at least k poles at the origin.
2.1 System Type
Poles at the origin s = 0 are known as free integrators. The System Type is dened as the
number of free integrators in the system.
(s z
1
) . . . (s z
m
)
Type 0: - G(s) = K
(s p
1
) . . . (s p
n
)
(s z
1
) . . . (s z
m
)
Type 1: - G(s) = K
s(s p
2
) . . . (s p
n
)
(s z
1
) . . . (s z
m
)
Type 2: - G(s) = K
s
2
(s p
1
) . . . (s p
n
)
and we can say
For a system under proportional control, to ensure e
ss
= 0 for a step input, the system
must be at least Type 1.
247
+
-
R(s)
C(s)
K + K
G (s)
PI control l er
pl ant
p
p
1
s
i
For a system under proportional control, to ensure e
ss
= 0 for a ramp input, the system
must be at least Type 2.
and we can make the following table showing the steady-state error conditions:
Type 0 Type 1 Type 2 Type 3
step
ramp
parabola
e
ss
= constant
e
ss
=
e
ss
e
ss
= 0
e
ss
= constant
e
ss
e
ss
= 0
e
ss
= 0
e
ss
= constant
e
ss
= 0
e
ss
= 0
e
ss
= 0 = =
Example 3
In the lab you (should have) observed that with proportional control (1) that the
velocity control gave a nite steady state error for a constant input, whereas (2)
the position control had zero steady-state error.
For velocity control:
(s) K
p
G(s) = =
d
Js + B
which is a Type 0 system, which will have a nite steady-state error. For position
control:
(s) K
p
G(s) = =
d
(s) s(Js + B)
which is a Type 1 system, and from the above argument will have a zero steady-
state error.
Example 4
Show why PI control reduces the steady-state error to zero for a step input with
a Type 0 system.
248
For a PI controller, the transfer function is
1 K
p
s + K
i
G
c
(s) = K
p
+ K
i
=
s s
so that the controller introduces (1) a pole at the origin, and(2) a zero at s =
K
i
/K
p
so that the forward transfer function is
(s + K
i
/K
p
)
G
f
(s) = G
c
(s)G
p
(s) = K
p
G
p
(s)
s
which is Type 1, and will have zero steady-state error for a constant input.
2.2 Static Error Constants
Recall that the transfer function relating the error to the input is
1
E(s) .
1 + G(s)
For a step input
1

1 1
e
ss
= lim s =
s0
s 1 + G(s) 1 + lim
s0
G(s)
If we dene a static position constant K
p
(not to be confused with a controller gain) as
1
K
p
= lim G(s) then e
ss
=
s0
1 + K
p
Similarly, for a ramp input (constant velocity)

1

1 1
e
ss
= lim s =
s0
s
2
1 + G(s) lim
s0
sG(s)
and, if we dene a static velocity constant K
v
as
1
K
v
= lim sG(s) then e
ss
= .
s 0
K
v
We can also dene an acceleration constant K

a
for parabolic inputs, since

1

1 1
e
ss
= lim s =
s0
s
3
1 + G(s) lim
s0
s
2
G(s)
so that if
Input Error Type 0 Type 1 Type 2
step
ramp
parabola
1/(1 + K
p
)
1/K
v
1/K
a
e
ss
= 1/(1 + K
p
)
K
v
= 0, e
ss
=
K
v
= 0, e
ss
K
p
= , e
ss
= 0
e
ss
= 1/K
v
K
v
= 0, e
ss
K
p
= , e
ss
= 0
K
v
= , e
ss
= 0
1/K
a
1
K
a
= lim s
2
G(s) then e
ss
= .
s 0
K
a
= =
249
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
gravi tati onal force acts to
move a di spl aced bal l away
from i ts equi l i bri um posi ti on
(b) an unstabl e system (a) a stabl e system (c) a neutral l y stabl e system
no forces act to move a
di spl aced bal l
gravi tati onal force acts to
restore a di spl aced bal l to
i ts equi l i bri um posi ti on

Spring Term 2008
Lecture 25
1
Reading:
Nise: Chap. 6
1 System Stability
The gure above illustrates three stability conditions using a rolling ball on an undulating
surface as a graphic example. Assume that the horizontal position and velocity of the ball
are a pair of state variables describing this system. In a concave region of the surface,
as shown in in (a) the base of the hollow is an equilibrium point. If the ball is displaced
a small distance from this position and released it oscillates but ultimately returns to its
rest position at the base as it loses energy due to frictional losses; this is therefore a stable
equilibrium point. (Without energy dissipation the ball would roll back and forth forever
and exhibit neutral, or marginal, stability.) On a convex portion of the surface, as shown in
(b), the ball is in equilibrium if placed exactly at the top of the surface, but if it is displaced
an innitesimal distance to either side the net gravitational force acting on the ball causes
it to roll down the surface and never to return to the equilibrium point. This equilibrium
point is therefore unstable. If the ball is displaced along the at portion of the surface, as
shown in (c) it neither moves away nor returns; the at portion represents a neutrally stable
equilibrium region.
There are many formal denitions of stability of a system.
From the solution of the linear dierential equation
y(t) = y
h
(t) + y
p
(t)
a system is dened to be asymptotically stable, if for all initial conditions, the steady-
state homogeneous response returns to zero. If the initial condition response increases
1
251
X
X
X
I
jM
s - pl ane
without bound the system is said to be unstable, and if the steady-state response is
constant or purely oscillatory the system is dened to be marginally stable.
From the forced response, a system is dened to be bounded-input bounded-output
(BIBO) stable only if every bounded input produces a bounded output.
These two denitions are not the same. A system
1
G(s) =
s
2
+
s
n
has a pair of imaginary poles at s = j
n
and will exhibit a purely oscillatory response from
any nite initial conditions. On this basis the system is therefore marginally stable. On
the other hand, if excited by an input sin(
n
t) the response amplitude will tend to innity,
therefore the system is BIBO unstable.
From our previous work on linear system response, we have seen that if any system pole
lies in the r.h. s-plane the system response will grow without bound as a result of any nite
initial condition on the output of its derivatives. A fundamental denition of system stability
is therefore
A linear system is dened to be stable only if all of its poles have negative real
parts. A system with one or more poles that are purely imaginary are dened to be
marginally stable.
The following gure shows a system with r.h. plane poles and the response components that
result.
Example 1
Determine the stability of the two systems
15 5
G
1
(s) = and G
2
(s) = .
s
3
+ 2s
2
+ 5s + 15 s
3
+ 2s
2
+ 5s + 5
252
MATLABs roots() function was used to nd the roots of the characteristic
equation for each system, to determine the system poles:
1. For G
1
(s): roots([1 2 5 15]) gave the system poles as s = 2.4357, s =
0.2268 j2.462.
2. For G
2
(s): roots([1 2 5 5]) gave the system poles as s = 1.233, s =
0.384 j1.977.
System G
1
(s) has a pair of complex conjugate poles in the r.h. plane and is
therefore unstable. System G
2
(s) has all of its poles in the l.h. plane and is
therefore stable.
The task of determining system stability is therefore that of determining the roots of the
characteristic equation D(s) = 0, or determining whether any root has a positive real part.
Before the advent of numerical software packages, such as MATLAB, this was a very dicult
problem.
2 The Routh-Hurwitz Criterion
The Routh-Hurwitz criterion is a method for determining the presence the number of roots
of a polynomial with positive real parts.
It is based on the coecients of the polynomial.
It involves placing the coecients into a 2-D Routh array, and then systematically
extending the array by creating new rows. When the array is complete, the number of
r.h. plane roots can be found by inspection.
The method has two basic steps:
Step 1: A preliminary step is the following
Given a polynomial, a sucient condition for the existence of at least
one r.h. plane root is that there is a sign change in the coecients.
For example the system
27
G(s) =
s
5
+ 47s
4
23s
3
+ 19s
2
+ 6s + 9
is unstable because there is a sign change in the coecients of D(s).
Put another way, for a system to be stable there must be no sign changes in the
coecients of the denominator polynomial D(s).
Step 2: The Routh array is created and examined, as described below.
253
2.1 The Routh Array
Let the denominator polynomial be
n
D(s) = a
n
s + a
n1
s
n1
+ . . . + a
1
s + a
0
the initial step is to place the coecients in the rst two rows of the array:
s
n
a
n
a
n2
a
n4
. . .
s
n1
a
n1
a
n3
a
n5
. . .
Additional rows are then created, based on the two rows above. For example, let the next
row be labelled s
n2
and s
n3
with entries b
i
and c
i
:
s
n
a
n
a
n2
a
n4
. . .
s
n1
a
n1
a
n3
a
n5
. . .
s
n2
b
1
b
2
b
3
. . .
s
n3
c
1
c
2
c
3
. . .
Each row is constructed from left to right, until all further entries are zero.
Each entry is based on entries in the two rows above, using values from (i) the rst
(leftmost) column, and the column to its right.
The entries us a 2 2 determinant.
In the above example the b
i
are computed from the rows labelled s
n
and s
n1
:
1
a
n
a
n2
a
n
a
n4
a
n
a
n6
b
1
=
a
n1
a
n1
a
n3
, b
2
=
a
n1
a
n1
a
n5
, b
3
=
a
n1
a
n1
a
n7
, . . .
and the c
i
are computed from the rows labelled s
n1
and s
n2
:
1 1 1
a
n1
a
n3
a
n1
a
n5
a
n1
a
n7
, c
1
= c
2
c
3
. . .
b
1
b
1
b
2
=
b
1
b
1
b
3
=
b
1
b1 b
4
The process is repeated until the row that will be labelled s
0
. Note that in each determinant
the left column is taken from the left column of the Routh array, the second column is taken
from the column to the right in the Routh array.
When the table is complete, the number of r.h. plane roots is the number of sign
changes in the rst column.
Example 2
Find the number of r.h. plane poles in the system
37
G(s) =
s
3
+ s
2
+ 2s + 24
254
The rst two rows of the Routh array are taken directly from D(s):
s
3
1 2 0
s
2
1 24 0
where an additional zero has been added at the end of each row. Then construct
the s
1
row:
b
1
=
1
a
n
a
n2
=
1
1
1
1
24
2
= 22
a
n1
a
n1
a
n3
1
a
n
a
n4
1 0
b
2
=
a
n1
a
n1
a
n5
=
1
1 0
= 0
and the Routh array, with the new row, becomes
s
3
1 2 0
s
2
1 24 0
s
1
-22 0 0
The s
0
row is now computed:
1
a
n1
a
n3
1 24
c
1
= 24 =
b
1
b
1
b
2
=
22
22 0
c
2
1
an 1 a
n3
1 0
= 0 =
b
1
b
1
b
3
=
22
22 0
and the completed Routh array is
s
3
1 2 0
s
2
1 24 0
s
1
-22 0
s
0
24 0
The next step is to examine the rst column; in this case we note there are
two sign changes 1 22 24, and we conclude that there are therefore two
unstable r.h. plane poles in this system.
In fact, for this system D(s) = (s + 3)(s 1 + j
7)(s 1 j
7).
Example 3
Use the Routh-Hurwitz method to develop a simple test for the stability of a
third-order system.
Let the denominator polynomial be
D(s) = a
3
s
3
+ a
2
s
2
+ a
1
s + a
0
255
and assume that all coecients are positive. The complete Routh array is
s
3
a
3
a
1
0
s
2
a
2
a
0
0
s
1
(a
3
a
0
a
2
a
1
)/a
1
0
s
0
a
0
0
and to ensure stability we require all terms in the rst column to be positive,
that is
a
1
a
2
> a
0
a
3
In other words a third-order system will be stable if the product of the inner
coecients is greater than the product of the outer coecients.
Note: The Routh-Hurwitz method has some special cases to deal with situations when
zeros appear in the rst column. See Nise, Chapter 6, for details.
256
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
+
-
R(s)
C(s)
P control l er
pl ant
K
1
s + 3s + 5s + 2
3 2

Spring Term 2008
Lecture 26
1
Reading:
Nise: Chapter 6
Nise: Chapter 8
1 Determining Stability Bounds in Closed-Loop Systems
Consider the closed-loop third-order system with proportional controller gain K with open-
loop transfer function
K
G
f
(s) =
s
3
+ 3s
2
+ 5s + 2
shown below.
The closed loop transfer function is:
N(s) K
G
cl
(s) = =
D(s) + N(s) s
3
+ 3s
2
+ 5s + (2 + K)
Lets examine the closed-loop stability by using the pzmap() function in MATLAB:
sys = tf(1,[1 3 5 2]);
pzmap(sys);
hold on;
for K = 2:2:30
sys = tf(K,[1 3 5 2+K]);
pzmap(sys);
end;
which superimposes the closed-loop pole/zero plots for K = 0 . . . 30 on a single plot:
1
261
PoleZero Map
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
4 3.5 3 2.5 2 1.5 1 0.5 0 0.5 1
3
2
1
0
1
2
3
K=0
K=30
K=30
K=30
K=0.
K=0
K increasing
K increasing
K increasing
From the plot we note the following:
This system always has two complex conjugate poles and a single real pole.
When K = 0 the poles are the open-loop poles.
As K increases, the real pole moves deeper into the l.h. plane, and the complex con
jugate poles approach and cross the imaginary (j) axis, and the system becomes
unstable.
Close examination of the plot shows that the system becomes unstable at a value of
K between K = 12 and K = 14.
We now look at three methods for determining the stability limit of the proportional gain
K for this system.
Example 1
Use the Routh-Hurwitz method to nd the range of proportional controller gain
K for which the above system will be stable.
The rst two rows of the Routh array are taken directly from D(s):
s
3
1 5 0
s
2
3 2 + K 0
262
and the next two rows are computed as above
1
a
n
a
n2
1 5
1
b
1
=
a
n1
a
n1
a
n3
= = (K 13)
3
3 2 + K
3
1
a
n
a
n4
1 0
b
2
= = 0
a
n1

a
n1
a
n5

=
3

3 0

Similarly, the s
0
row is computed
1
a
n1
a
n3
3 24
c
1
= = 2 + K
b
1
b
1
b
2
=
K 13
(K 13)/3 2 + K
1
a
n1
a
n3
3 0
c
2
= = 0
b
1
b
1
b
3
=
K 13
(K 13)/3 0
and the complete Routh array is
s
3
1 5 0
s
2
3 2+K 0
s
1
-(K-13)/3 0
s
0
(2+K) 0
We now examine the rst column to determine the range of proportional gain
for which this system will be stable. In order for there to be no sign changes we
require
2 < K < 13
We conclude that if K < 2 there will be one (therefore real) unstable pole,
while if K > 13 there will be two unstable poles. When K = 13 the denominator
is
D(s) = s
3
+ 3s
2
+ 5s + 15 = (s + 3)(s + j2.236)(s j2.236)
so that the closed-loop system has a pair of poles on the imaginary axis. The
system will be marginally stable (a pure oscillator at a frequency of = 2.236
r/s).
Example 2
Use the stability criterion for third-order systems developed in Example 3 of
Lecture 25 to determine the stability bounds for the above system.
In Lecture 25 we showed that for a third-order system with characteristic equa
tion:
D(s) = a
3
s
3
+ a
2
s
2
+ a
1
s + a
0
= 0
the system is stable only if
a
1
a
2
> a
0
a
3
263
In this case
D(s) = s
3
+ 3s
2
+ 5s + (2 + K)
and therefore for stability we require
15 > 2 + K
or K < 13.
Example 3
Use the characteristic equation directly to nd the closed-loop stability limits for
the above system. There are three closed-poles. We conjecture that at the stabil
ity boundary (marginal stability) there will be a pair of poles on the imaginary
axis at s = j, and a single real pole at s = a.
The closed-loop characteristic polynomial will therefore be
D(s) = (s + a)(s
2
+
2
) = s
3
+ as
2
+
2
s + a
2
Comparing coecients with the actual closed-loop characteristic polynomial
D(s) = s
3
+ 3s
2
+ 5s + (2 + K)
we determine
a = 3
2
= 5 =
5
a
2
= K + 2 K = 13
2 Root Locus Methods
We have seen that the closed-loop poles change as controller parameters vary. A root-locus
is is an s-plane plot of the paths that the closed-loop poles take as a controller parameter
varies. Lets start with some simple examples.
Example 4
Consider the rst order plant under proportional control, as shown below:
264
x I
jM
s-pl ane
K=0
-a
K increasing
the root l ocus
i ndi cates the path
of the pol e as K
vari es.
+
-
R(s) C(s)
P-D control l er
pl ant
K + K s
1
s + a
p d
+
-
R(s) C(s)
P control l er
pl ant
K
1
s + a
K
G
cl
(s) =
s + (a + K)
with a single pole p
c]
= (a + K). The root-locus is simply the path of this pole
as K varies from K = 0 to K = Clearly as K 0, the closed-loop pole .
approaches the open-loop pole (s = a), and as K , the closed-loop pole
p . This is all the information we need to construct the root-locus for this
system.
Example 5
Construct the root-locus plot for the rst-order system under P-D control with
G
c
(s) = K
p
+ K
d
s:
265
x
I
jM
s-pl ane
K=0
-a
K increasing
-b
o
x
I
jM
s-pl ane
K=0
-a
K increasing
-b
o
(a) b> a (b) a> b
If we write

K
p
G
c
(s) = K
d
s +
K
d
we have a open-loop pole at s = a and an open-loop zero at s = K
p
/K
d
= b.
K
d
(s + b)
G
cl
(s) =
(K
d
+ 1)s + (a + K
d
b)
with a single pole
a + bK
d
p
cl
. =
1 + K
d
We now construct the root-locus as K
d
varies from 0 to . Clearly as K
d
0,
the closed-loop pole p
cl
a approaches the open-loop pole at (s = a), and
as K
d
, the closed-loop pole p
cl
b, in other words the closed-loop pole
approaches the open-loop zero. There are two possibilities for the root locus
based on the relative positions of the open-loop pole and zero:
While the root locus always originates at the pole and terminates at the zero, if
b > a the closed-loop pole will move to the left, while if a > b the pole will move
to the right. In addition we can note:
There is a closed-loop zero at s = b.
This is a case when the order of the numerator is equal to the order of the
denominator, and there will be direct feed-through from the input to the
output, as discussed in Lecture 22.
As K is increased, and the closed-loop pole approaches the zero, the strength
of the component e
p
cl
t
in the response will be diminished (Lecture 23).
266
+
-
R(s)
C(s)
P control l er
pl ant
K
w
s + 2zw s + w
2
n
n
2
n
2
Example 6
Determine the root locus for the second-order system
2
G(s) =
n
s
2
+ 2
n
s +
n
2
under proportional control.
K
2
G(s) =
n
s
2
+ 2
n
s +
n
2
(1 + K)
with closed-loop poles
p
1
, p
2
=
n

n
2
(1 + K)
which will be real only if 1 and K 1, otherwise they will be complex
conjugates. We note the following:
As
2
0, the closed-loop poles approach the open-loop poles
n
n
1.
If the open-loop poles are real, the closed-loop poles will move together as
K
2
1, and then become complex.
If the closed-loop poles are complex, p
1
, p
2
=
n
j
n
(1 + K)
2
,
only the imaginary part is aected by K, and as K the closed-loop
poles p
1
, p
2
n
j.
This behavior is summarized in the following root locus plots:
267
x s
jw
s-pl ane
K = 0
K i ncreasi ng
(a) z > 1
x
K i ncreasi ng
K = 0
K = z - 1
x
s
jw
s-pl ane K = 0
K i ncreasi ng
(b) z < 1
x
K i ncreasi ng
K = 0
2.1 Some Basic Properties of Root Locus Plots
2.1.1 The Number of Branches in the Plot
By denition there will be one branch of the plot for each closed-loop pole. For a system
with open-loop transfer function
N
ol
(s)
G
ol
(s) = K
D
ol
(s)
The closed-loop characteristic polynomial is
D
cl
(s) = D
ol
(s) + KN
ol
(s)
and provided the order of N
ol
(s) does not exceed that of D
ol
(s), the order of D
cl
(s) will
be the same as that of D
ol
(s). In other words, the number of closed-loop poles equals the
number of open-loop poles, and the number of branches equals the number of open-loop
poles..
2.1.2 Symmetry of the Root Locus Plot
Because all closed-loop poles are either real or complex conjugate pairs, the root locus is
symmetrical about the real axis. The implication of this is that when we discuss rules for
generating a root locus, we only have to consider half of the s-plane.
2.1.3 The Origins of the Branches (K = 0)
The closed-loop characteristic polynomial is
D
cl
(s) = D
ol
(s) + KN
ol
(s).
As K 0, D
cl
(s) D
ol
(s), with the result that the n branches of the root locus always
originate at the open-loop poles.
268
2.1.4 The Terminal Points of the Branches (K )
As K becomes large
D
cl
(s) KN
ol
(s)
with the result that m of the n closed-loop roots approach the m open-loop zeros. This
leaves n m roots to be accounted for, and we will investigate this later. For now we simply
state that these branches diverge away from the origin along a set of n m straight-line
asymptotes, and as K these poles approach a distance r = from the origin.
269
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
+
-
R(s)
C(s)
KG (s)
G (s)
control l er
pl ant
c
p
H(s)

Spring Term 2008
Lecture 27
1
Reading:
Nise: Chapter 8
1 Root Locus Development
In Lecture 26 we saw some simple examples of root locus plots. We now look at the general
method of generating root loci.
Recall that the closed-loop characteristic equation is:
1 + KG(s) = 0
where G(s) = G
c
(s)G
p
(s)H(s) is the open-loop transfer function. We now ask ourselves:
how can we tell if an arbitrary point s = + j lies on the root locus? In other words
we seek conditions that determine whether s is a root of the characteristic equation. From
above, s is a root if
KG(s) = 1 + j0.
In polar form this may be expressed as
KG(s) = KG(s) e
j
(KG(s))
|
j(2n
|
+1)
= 1 e for n = 0, 1, 2, . . .
= cos((2n + 1)) + j sin((2n + 1))
= 1 + j0.
This tells us that for any point s = + j on the root locus
|KG(s)| = 1 and
(G(s)) = (2n + 1)
which generates two important conditions:
1
271
| |
The Angle Condition:
(G(s)) = (2n + 1)
The Magnitude Condition: KG(s) = 1 | |
In practice, the angle condition is used to determine whether a point s lies on the root locus,
and if it does, the magnitude condition is used to determine the gain K associated with that
point, since K = 1/ G(s) . | |
Example 1
Given the open-loop system
1
G(s) =
(s + 2)(s + 4)
determine whether the points s = 1, s = 3.5, s = 3 + j5 are on the root
locus.
For s = 1:
K K
KG(s) =
(1 + 2)(1 + 4) 3
KG(s) = 1 for any K > 0, so we conclude s = 1 is not on the root-locus.
For s = 3.5:
K K
KG(s) =
(3.5 + 2)(3.5 + 4) 0.75
KG(s) = 1 for K = 0.75, so we conclude s = 3.5 lies on the root-locus.
For s = 3 + j5:
K K K
KG(s) = = =
((3 + j5) + 2)((3 + j5) + 4) (1 + j5)(1 + j5) 26
KG(s) = 1 for K = 26, so we conclude s = 3 + j5 lies on the root-locus.
1.1 Geometric Evaluation of the Transfer Function
The transfer function may be evaluated for any value of s = + j, and in general, when s
is complex the function G(s) itself is complex. It is common to express the complex value
of the transfer function in polar form as a magnitude and an angle:
G(s) = G(s) e
j
G(s)
, | |
with a magnitude G(s) and an angle G(s) given by
|G(s)| =

{G(s)}
2
+ {G(s)}
2
,
G(s) = tan
1
{G(s)}

{G(s)}
272
0
s-pl ane
s
p
1
2
(s - p )
|
s
|
2
p
2
jw
s
where {} is the real operator, and {} is the imaginary operator.

If the numerator and denominator polynomials are factored into terms (sp
i
) and (sz
i
),
G(s) = C
(s z
1
)(s z
2
) . . . (s z
m1
)(s z
m
)
,
(s p
1
)(s p
2
) . . . (s p
n1
)(s p
n
)
(where C is a constant), each of the factors in the numerator and denominator is a complex
quantity, and may be interpreted as a vector in the s-plane, originating from the point z
i
or
p
i
and directed to the point s at which the function is to be evaluated. Each of these vectors
may be written in polar form in terms of a magnitude and an angle, for example for a pole
p
i
=
i
+ j
i
, the magnitude and angle of the vector to the point s = + j are
|s p
i
| =
(
i
)
2
+ (
i
)
2
,
(s p
i
) = tan
1

i

i
as shown below.
Because the magnitude of the product of two complex quantities is the product of the
individual magnitudes, and the angle of the product is the sum of the component angles,
that is if a and b are complex, that is
|ab| = |a| |b| , |a/b| = |a| / |b|
(ab) =
a +
b,
(a/b) =
b
the magnitude and angle of the complete transfer function may then be written
m
i=1
(s z
i
)
|G(s)| = C
i
n
=1
|
(s p
i
)
|
m
| |
n
G(s) =

(s z
i
)

(s p
i
).
i=1 i=1
The magnitude of each of the component vectors in the numerator and denominator is the
distance of the point s from the pole or zero on the s-plane. Therefore if the vector from the
pole p
i
to the point s on a pole-zero plot has a length q
i
and an angle
i
from the horizontal,
273
0
s-pl ane
s
p
1
p
2
z
1
q
q
1
2
f
1
q
1
2
1
q
r
jw
s
G(s) = j - q - q
1
1 2
| G(s)| = C
r
q q
1
1 2
and the vector from the zero z
i
to the point s has a length r
i
and an angle
i
, as shown
above, the value of the transfer function at the point s is
r
1
. . . r
m
|G(s)| = C
q
1
. . . q
n
G(s) = (
1
+ . . . +
m
) (
1
+ . . . +
n
)
The transfer function at any value of s may therefore be determined geometrically from the
pole-zero plot, except for the overall gain factor C. The magnitude of the transfer function
is proportional to the product of the geometric distances on the s-plane from each zero to
the point s divided by the product of the distances from each pole to the point. The angle of
the transfer function is the sum of the angles of the vectors associated with the zeros minus
the sum of the angles of the vectors associated with the poles.
The angle condition then states that for a point s = + j to be on the root locus,
m n
G(s) =

i
= (2n + 1),
i=1 i=1
and once it has been established that s lies on the locus, the magnitude condition may be
used to determine the value of K:
1
K =
|G(s)|
Example 2
For open-loop system
s + b
G(s) =
s + a
use the angle condition to determine whether the points labeled A, B, C and D
lie on the root locus.
274
x
P
s
jw
o
q = 0
f = 0
x
P
s
jw
o
q = p
f = p
x
-a
A
B C
I
jM
G
o
B
D
-b
For an arbitrary point s
G(s) = .
At Point A: = (2 n + 1), therefore A is not on the root locus.
At Point B: = 0, = 0, = 0, therefore B is not on the root locus.
At Point C: = , = 0, = , therefore C is on the root locus.
At Point D: = , = , = 0, therefore D is not on the root locus.
The only one of these four points that satises the angle condition, and therefore
lies on the root locus, is point C.
1.2 Regions of the Real Axis on the Root Locus
We note that for a point s = that lies on the real axis:
(a) Poles and zeros on the real axis that lie to the left of the point s contribute zero to the
angle condition.
(b) Poles and zeros on the real axis that lie to the right of the point s contribute to the
angle condition.
275
N
N
q
2
f = 2p - f
1
1 2
s
jw
x
x
I
jM
x
x o o o
1 pol e/zero to
the ri ght
3 pol es/zeros to
the ri ght
5 pol es/zeros to
the ri ght
7 pol es/zeros to
the ri ght
(c) A complex conjugate pole or zero pair contributes a total of zero (2)to the angle
condition along the real axis.
These observations combine to generate the condition:
A point on the real axis lies on the root locus only if there are an odd
number of poles and/or zeros to its right.
Example 3
Dene the regions of the real axis that will lie on the root locus for the following
open-loop pole-zero plot with 4 poles and 3 zeros, and then qualitatively ll in
the rest of the plot.
The real axis regions are shown below:
276
x
x
s
jw
x
x o o o
K=0
K=0
K=0
K=0 K= K=
K=
K
Since we know that (1) branches must originate from open-loop poles, and (2)
terminate on open-loop zeros or go to innity, we have enough information to
sketch the form of the root locus:
where we note that branches must originate from the complex conjugate pole
pair and terminate on the r.h. plane zeros.
1.3 Behavior of the Root Locus for Large Values of K
Let the closed-loop transfer function be
KG(s) KN(s)
G
cl
(s) = =
1 + KG(s) D(s) + KN(s)
where N(s) is of order m, and D(s) is of order n. We can see from the above expression
that provided n m the closed-loop system will also be of order n. When K becomes large
we can approximate the closed-loop characteristic equation as
KN(s) = 0
(which is of order m) and so we can state:
As the value of K , m of the closed-loop poles approach the m
open loop zeros.
This leaves n m closed-loop poles unaccounted for. Lets assume that for large K, where
m of the poles are very close to the zeros, pole-zero cancellation has taken place and the
characteristic equation becomes
K K
1 + KG(s) 1 +
(s p
i
)
1 + = 0
i
s
nm
277
n-m = 1
180
o
n-m = 2
270
o
90
o
180
o
300
o
60
o
n-m = 3
n-m = 4
45
o 135
o
225
o
315
o
where the p
i
are the n m uncancelled poles. With this approximation
s
nm
= K, or s = K
1/(nm)
(1)
1/(nm)
The n m roots of 1 are complex with values
s
k
= e
j(2k+1)/(nm)
k = 0, 1, . . . n m 1
that is, they lie equally spaced around the unit circle at angles
(2k + 1)
k
= , k = 0, 1, . . . n m 1.
n m
and as K becomes large, the n m closed-loop poles approach a set of radial asymptotes at
these angles. The asymptotic angles are summarized in the following table:
n m Asymptote Angles
1 180

2 90
, 270

3 60
, 180
, 300

4 45
, 135
, 225
, 315

As the gain K becomes large, n m branches of the root locus diverge
away from the origin and approach n m radial asymptotes, at angles
k
= (2k + 1)/(n m), for k = 0 . . . (n m 1).
This is not quite the full picture. We will investigate this further in the next lecture.
278
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
+
-
R(s)
C(s)
KG (s)
G (s)
control l er
pl ant
c
p
H(s)
Spring Term 2008
Lecture 28
1
Reading:
Nise: Chapter 8
1 Root Locus Development (contd. from Lecture 27)
1.1 Behavior of the Root Locus as the Gain K Becomes Large (contd.)
For a closed-loop system
with open-loop transfer function G(s) = KG
c
(s)G
p
(s)H(s), we saw in Lecture 27 that the
asymptotic angles are summarized in the following table:
n m
1
2
3
4
Asymptote Angles
180

90
, 270

60
, 180
, 300

45
, 135
, 225
, 315

1
281
n-m = 1
180
o
n-m = 2
270
o
90
o
180
o
300
o
60
o
n-m = 3 n-m = 4
45
o 135
o
225
o
315
o
We now rene this property a little further. Let
s
m
+ b
m1
s
m1
+ b
m1
s
m2
+ . . . + b
0
(s z
1
)(s z
2
) . . . (s z
m
)
G(s) = K = K .
s
n
+ a
n1
s
n1
+ a
n1
s
n2
+ . . . + a
0
(s p
1
)(s p
2
) . . . (s p
n
)
For any polynomial of degree k, the coecient of the term in s
k1
is the sum of the roots of
the polynomial, so that
b
m1
= (z
1
+ z
2
+ . . . + z
m
)
a
n1
= (p
1
+ p
2
+ . . . + p
n
).
Further for large s, assume that we can ignore all but the rst two terms in each polynomial,
and write:
s
m
+ b
m1
s
m1
s
m
(z
1
+ z
2
+ . . . + z
m
)s
m1
G(s) K
s
n
+ a
n1
s
n1
= K
s
n
(p
1
+ p
2
+ . . . + p
n
)s
n1
.
For large K, m closed-loop poles approach m of the open-loop zeros (which are also closed-
loop zeros). and if we assume pole/zero cancellation we can do polynomial division and
write
1
G(s) K
s
nm
+ (a
n1
b
m1
)s
nm1
We can then use the relationship that for large x
x
k
+ cx
k1
(x + c/k)
k
to write
1
G(s) K
(s
a
)
nm
where
a
=
a
n1
b
m1
n m
=
(p
1
+ p
2
+ . . . + p
n
) (z
1
+ z
2
+ . . . + z
m
)
n m
282
n-m = 3
n-m = 4
I
I
jM
jM
I I
a a
60
o
45
o
The closed-loop characteristic equation for large s and large K may be approximated
K
1 + KG(s) 1 +
(s
a
)
nm
= 0
and the asymptotes are given by
(s
a
) = K
1/(nm)
(1)
1/(nm)
which is similar to the result obtained in Lecture 27, namely that the roots lie on a circle of
radius K
1/nm
, but this time with the dierence that the asymptotes radiate from the point
s =
a
. The point s =
a
is dened as the centroid of the asymptotes and

poles

zeros
a
= ,
n m
and
a
is always real because poles and zeros are either real of complex conjugates. The
following gure shows the displacement of the origin of the asymptotes:
The root locus sketching rule, for constructing the locus of a system with m open-loop zeros,
and n open-loop poles, for large values of gain K is therefore:
As the value of K , m of the closed-loop poles approach the
m open loop zeros.
As the gain K becomes large, n m branches of the root locus
diverge away from a point s =
a
on the real axis and approach
n m radial asymptotes, at angles
k
= (2k + 1)/(n m), for k =
0 . . . (n m 1).
Example 1
Sketch the root locus plot for the open-loop system
1
G(s) = K
(s + 1)(s + 2)(s + 4)
and nd the gain K at which it becomes unstable. We proceed as follows:
283
x x x
-4 -2
-1
60
-60
-7
s = - --
3
jw
a
1. Determine and plot the open-loop poles and zeros.
2. Determine and plot the regions of the real axis that lie on the root locus.
3. Determine the number of asymptotes. There are no nite zeros, therefore
n m = 3.
4. Determine the asymptote angles and centroid, then sketch the asymptotes.
For three asymptotes the angles are (see the above table) 60
, 180
, 300
.
The centroid is
a
= ((sum of the poles) (sum of the zeros))/(n m)
1 7
= ((1 2 4) (0)) =
3 3
These steps were used to produce the following sketch:
The closed-loop characteristic equation is:
(s + 1)(s + 2)(s + 4) + K = s
3
+ 7s
2
+ 14s + 8 + K = 0
and at the point of marginal stability (when s = j)
j
3
7
2
+ j14 + 8 + K = 0 + j0
Equating the real and imaginary parts
7
2
+ 8 + K = 0
3
+ 14 = 0
284
giving
= 0,
14
K = 8, 90
Since the roor locus is dened only for K > 0 we conclude that the system will
become unstable for K > 90, and the locus will cross the imaginary axis at
s = j
(14) rad/s.
Example 2
Show the eect of PD control on the root-locus of the previous example.
Let
G
c
(s) = K
p
+ K
d
s = K(s + b),
where b = K
p
/K
d
and K = K
d
. The PD controller has added a zero at s = b
to the system. The open-loop transfer function is now
s + b
G(s) = K
(s + 1)(s + 2)(s + 4)
and we have n m = 2.
Assume for now that b = 3. There will be n m = 2 asymptotes, at angles 90

and 270
. The centroid will be

1
a
((1 2 4) (3)) = 2
2
Now assume b = 6. The centroid will be
1
a
((1 2 4) (6)) = 0.5
2
These two cases are sketched below:
285
x x x
-2
-1
90
0
s = - 2
s
jw
a
o
-3 -4
x x x
-2
-1
90
0
s = - 0.5
s
jw
a
o
-6 -4
b = 6
b = 3
(a) P-D zero at s= -3 (b) P-D zero at s= -6
Notice that as the PD zero moves deeper into the l.h. plane, it moves the asymp
tote toward the imaginary axis, meaning that the dominant closed-loop poles
become lightly damped. You can show for yourself that if b > 7, the asymptote
origin
a
> 0 and this system will become unstable as K is increased.
Example 3
Show some typical eects of PID control on the root-locus of the previous exam
ple.
With PID control
K
i
G
c
(s) = K
p
+ + K
d
s
s
s
2
+ (K
p
/K
d
)s + K
i
/K
d
= K
d
s
(s z
1
)(s z
2
)
= K
s
The PID controller has added a pole at the origin, and two zeros, which we as
designers can place in order to shape the root locus to meet a set of specications
on the dynamic response. The open-loop transfer function is now
G(s) = K
(s z
1
)(s z
2
)
s(s + 1)(s + 2)(s + 4)
286
x x x
-2
-1
90
0
s = 1.25
s
jw
a
o
-4. 5
x x x
-2
-1
90
0
s = - 1.5
jw
a
o
-4
x
o
-4 -5
o
x
-j2
j2
and we have n m = 2. There will be n m = 2 asymptotes, at angles 90

and
270
. The two root loci below show cases for

(a) two real zeros (z
1,2
= 4.5, 5), and
(b) two complex conjugate zeros (z
1,2
= 2 j2).
In the rst case the centroid will be
1
a
= ((0 1 2 4) (4.5 5)) = 1.25,
2
while in the second case the centroid will be
1
a
= ((0 1 2 4) (2 + j2 2 j2)) = 0.5
2
These two cases are sketched below:
287
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
X X o o
a "breakaway" poi nt
a "break-i n" poi nt
s
jw
K = 0 K = 0
K i ncreasi ng
K i ncreasi ng
K = K =

Spring Term 2008
Lecture 29
1
Reading:
Nise: Chapter 8
1 Root Locus Renement
The complete set of sketching rules contains additional methods to make a sketched plot
more accurate. While these were useful in the days before ubiquitous computation, today
with the existence of tools such as MATLAB makes these graphical renements somewhat
unnecessary. We therefore just mention them here and refer you to Nise, Section 8.5, for
more detail.
1.1 Real-Axis Breakaway and Break-In Points
A breakaway point is the point on a real axis segment of the root locus between two real
poles where the two real closed-loop poles meet and diverge to become complex conjugates.
Similarly, a break-in point is the point on a real axis segment of the root locus between two
real zeros where two real closed-loop complex conjugate zeros meet and diverge to become
real.
Because the closed-loop poles originate from open-loop poles (when K = 0), a breakaway
point will correspond to the point of maximum K along the real-axis segment. Similarly, a
break-in point will correspond to the point of minimum K on the real axis segment between
the two zeros.
The closed-loop characteristic equations is 1 + KG(s) = 0, so that along the root locus
segments on the real axis (s = )
1 D()
K =
G()
=
N()
1
291
X X o
o
"breakaway" poi nt
a"break-i n" poi nt
I
jM
-1
-2 -4 -6
X X o o
s
K = 0 K = 0
K = K =
K
K
s s
0 0
s s
b b
The breakaway/break-in points (maximum/minimum points) will therefore occur where
dK d

D()

d
=
dK N()
= 0
or when
N()D
() N
()D() = 0.
Example 1
Find the real axis breakaway/break-in points for the closed-loop system with
s
2
+ 10s + 24 (s + 6)(s + 4)
G(s) = = .
s
2
+ 3s + 2 (s + 1)(s + 2)
The root locus has two real-axis segments, between the pole pair and between
the zero pair. There will therefore be a breakaway point and a break-in point.
The breakaway/break-n points will be contained in the roots of
N()D
() N
()D() = 0.
or
(
2
+ 10 + 24)(2 + 3) (
2
+ 3 + 2)(2 + 10) = 7
2
+ 44 + 52 = 0
giving
1,2
= 4.708, 1.578, as shown below:
292
I
jM
o
X
G
B
angl e of arri val
angl e of departure
I
jM
o
X
G
B
angl e of departure
o
X
G
B
A
1
d
1 2
X X o
o
s = -1. 578 s = -4. 708
s
jw
-1
-2 -4 -6
1.2 Angle of Arrival and Departure from Zeros and Poles
Further renement of the Root Locus may be made by computing the angle at which the
branches of the locus depart from the open-loop poles, and arrive at the open-loop zeros.
Consider a point a small distance from a pole:
The angle condition at the point requires

angles from the zeros

angles from the poles = (2k + 1)
or
1
+
2

1

d
= (2k + 1)
293
I
jM
X
o
X
G
B
d
j2
angl e of departure
-j2
-1
-4
G
-1 0 -8 -6 -4 -2
0
-4
-2
2
4
123. 7
o
I
jM
-7. 61
Let k = 0 and let 0, then
d
=
1
+
2

1

where the angles are measured to the pole itself.
A similar argument denes the angle of arrival at a complex zero.
Example 2
Find the angle of departure at the pole p = 1 + j2 for the closed-loop system
where
s + 4
G(s) = .
s
2
+ 2s + 5
In the above gure = 90
, = arctan(2/3) = 33.7
. The angle of departure

is therefore
d
=
= 33.7

90

180

= 236.31

= 123.69

294
1.3 Summary of Root Locus Sketching Rules
Denitions The open-loop transfer function is KG
c
(s)G
p
H(s) which can
be rewritten as KN(s)/D(s).
N(s), the numerator, is an mth order polynomial; D(s) is nth
order.
G(s) has zeros at z
i
, (i = 1 . . . m); and poles at p
i
(i = 1 . . . n).
Symmetry The locus is symmetric about real axis (i.e., complex poles ap
pear as conjugate pairs).
Number of branches There are n branches of the locus, one for each pole of the closed-
loop transfer function.
Start and end points The locus starts (when K = 0) at poles of the open-loop transfer
function, and ends (when K = ) at the zeros. Note: there are
n m zeros of the open-loop transfer function as |s| .
Locus on real axis The locus exists on the real axis to the left of an odd number of
poles and zeros.
Asymptotes as |s| If n > m there are n m asymptotes of the root locus that
intersect the real axis at
a
= (

p
i

z
i
)/(nm), and radiate
out with angles
k
= (2k +1)/(nm), for k = 0 . . . (nm1).
Renement of the Root Locus:
Breakaway and break-in
points on the real axis
There are breakaway or in points of the locus on the real axis
wherever N(s)D
(s) N
(s)D(s) = 0.
Angle of departure from
a complex pole
The angle of departure from pole p
j
is
d,p
j
= 180

+
m

i=1
(p
j
z
i
)
n

i=1,i=j
(p
j
p
i
)
Angle of arrival at com
plex zero
The angle of arrival at zero z
j
is
a,z
j
= 180

+
m

i=1,i=j
(z
j
z
i
)
n

i=1
(z
j
p
i
)
Imaginary axis crossings
(stability limits)
Use Routh-Hurwitz to determine where the locus crosses the
imaginary axis, or assume a form for the closed-loop char. eqn.
and solve for the coecients
Determine the poles for
a given gain K
Substitute the value of K into D(s) + KN(s) = 0 and nd roots
of characteristic equation. (This may require a computer)
Determine K for a given
pole location
Use the magnitude condition with s = + j, ie K =
D(s)/N(s). (If s is not exactly on the locus, K may be com
plex, but the imaginary part should be small. Take the real part
of K for your answer.)
295
Root Locus
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
12 10 8 6 4 2 0 2
2
1.5
1
0.5
0
0.5
1
1.5
2
0.92 0.95
0.98
0.92 0.95
0.98
3 4 5 6 7
2
2.1
rlocus()
For example
sys =
rlocus(sys)
produces the plot
MATLAB Root Locus Functions
MATLAB Language Functions
The function rlocus(sys) produces a root locus plot for the system object sys.
zpk([ -1.5 -4.5],[0 -1 -4],1)
12 10 8 6 4 2 0 2
2
1.5
1
0.5
0
0.5
1
1.5
2
Root Locus
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
sgrid() The function sgrid without any arguments generates a grid over an existing con
tinuous s-plane root locus or pole-zero plot. Lines of constant damping ratio () and
undamped natural frequency (
n
) are drawn.
If invoked with a pair of arguments, sgrid(zeta, wn), lines of constant damping ratio
for the values given in the vectors zeta and wn will be plotted. For example
sgrid([0.92 0.95 0.98], [3 4 5 6 7]) superimposed on the above example pro
duces the plot
296
Root Locus
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
7 6 5 4 3 2 1 0 1
2.5
2
1.5
1
0.5
0
0.5
1
1.5
2
2.5
System: sys
Gain: 4.52
Pole: 1.97 + 1.97i
Damping: 0.707
Overshoot (%): 4.31
Frequency (rad/sec): 2.78
0.707
0.707
Example 3
Use MATLAB to nd the gain K to achieve a closed-loop damping ratio 0f
= 0.707 for the open-loop system
(s + 1)
2
G(s)
s(s
2
+ 1)
The commands sys = tf([1 2 1],[1 0 1 0])
sgrid(0.707, 0)
produce the plot
With the cursor, the gain is found to be K = 4.52. The step response of the
closed-loop system is found with the following commands:
closed loop = feedback (4.52*sys, 1)
step(closed loop)
297
14 12 10 8 6 4 2 0 2
2
1.5
1
0.5
0
0.5
1
1.5
2
Root Locus
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
0 1 2 3 4 5 6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
We note that the overshoot is greater than the 5% predicted by the poles with
a damping ratio of = 0.707. This is because the closed-loop system retains the
open-loop zeros, and these zeros accentuate the overshoot.
Example 4
Plot the root locus for
(s + 1.5)(s + 5.5)
G(s) .
s(s + 1)(s + 5)
sys = zpk([-1.5 -5.5],[0 -1 -5],1)
rlocus(sys)
298
2.2 RLTOOL, an Interactive Root Locus Design Tool
RLTOOL is a variant of SISOTOOL, which is an interactive GUI based control system design
tool in MATLAB. It will be introduced through a classroom demo.
The plant G
p
(s) may be imported as a system object from the MATLAB workspace.
The compensator/controller may be dened interactively by placing poles and zeros
directly on the root locus plot.
These poles may be moved around by dragging on the plot.
Response curves (step, impulse, etc) may be displayed as the gain is changed.
299
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
A
-A
0
T =
t
A si n(B)
A si n(Mt + B)
2F
M

Spring Term 2008
Lecture 30
1
Reading:
Nise: 10.1
1 Sinusoidal Frequency Response
1.1 Denitions
Consider a sinusoidal waveform
f(t) = A sin (t + )
where
A is the amplitude (in appropriate units)
is the angular frequency (rad/s)
is the phase (rad)
In addition we can dene
T the period T = 2/ (s)
f the frequency, (f = 1/T = /2) (Hz)
1
301
F(t)
0 t
I nput
t
o
mass
m
F(t)
fri cti on B
Response
steady-state
response
transi ent
response
v (t)
m
v (t)
m
The Euler Formulas: We will frequently need the Euler formulas
e
jt
= cos (t) + j sin (t)
e
jt
= cos (t) j sin (t)
or conversely
cos (t) =
1
e
jt
+ e
jt

2
sin (t) =
1
e
jt
e
jt

2j
1.2 The Steady-State Sinusoidal Response
Assume a system, such as shown above, is excited by a sinusoidal input. The total response
will have two components a) a transient component, and a steady-state component
y(t) = y
h
(t) + y
p
(t).
We dene the steady-state component as the particular solution y
p
(t). Let the system dif
ferential equation be
d
n
y d
n1
y dy d
m
u d
m1
u du
a
n
dt
n
+ a
n1
dt
n1
+ . . . + a
1
dt
+ a
0
y = b
m
dt
m
+ b
m1
dt
m1
+ . . . + b
1
dt
+ b
0
u
with a complex exponential input
u(t) = e
jt
.
Assume a particular solution y
p
(t) to be of the same form as the input, that is
y
p
(t) = Ae
jt
and since
d
k
y
p
= A(j)
k
e
jt
dt
k
substitution into the dierential equation gives:
a
n
(j)
n
+ a
n1
(j)
n1
+ . . . + a
1
(j) + a
0

Ae
jt
=
b
m
(j)
m
+ b
m1
(j)
m1
+ . . . + (b
1
j) + b
0

e
jt
302
H(s)
u(t) = e y (t) = H(jM)e
jMt jMt
ss
Li near System
or
b
m
(j)
m
+ b
m1
(j)
m1
+ . . . + b
1
(j) + b
0
A =
a
n
(j)
n
+ a
n1
(j)
n1
+ . . . + a
1
(j) + a
0
Examination of this equation shows its similarity to the transfer function H(s), in fact
A = H(s)
s=j
= H(j) |
so that the steady-state response y
ss
(t) is
y
ss
(t) = y
p
(t) = Ae
jt
= H(j)e
jt
, (1)
or in other words, the steady-state response to a complex exponential input is dened by
the transfer function evaluated at s = j, or along the imaginary axis of the s-plane. Note
that H(j) is in general complex.
We now extend this argument to a real sinusoidal input, for example u(t) = cos (t) =
(e
jt
+ e
jt
)/2. The principle of superposition for linear systems allows us to express the
response as the sum of the two responses to the complex exponentials:
y
ss
(t) =
1
H(j)e
jt
+ H(j)e
jt

2
We now proceed as follows:
We show that H(j) = H(j) where H(j) denotes the complex conjugate (see the
Appendix), so that
y
ss
(t) =
1
H(j)e
jt
+ H(j)e
jt

(2)
2
We break up H(j) into its real and imaginary parts,
H(j) =
H(j) =
and use the Euler formula to write
{H(j)} + j {H(j)}
{H(j)} j {H(j)}
e
jt
e
jt
=
=
cos (t) + j sin (t)
cos (t) j sin (t)
We combine the real and imaginary parts of Eq. (2) to conclude
y
ss
(t) = {H(j)} cos(t) {H(j)} sin(t) (3)
303
| u(t)|
| y(t)|
T
t
u(t) y(t)
Ti me
peri od
Dt
0
H(jw) = -
| H(jw)| =
ampl i tude rati o:
| y(t)|
| u(t)|
phase shi ft: 2p
Dt
T
A
m
p
l
i
t
u
d
e
We then use the trig. identity

a cos b sin =
a
2
+ b
2
cos( + )
to write Eq. (3) as
y
ss
(t) = H(j) cos (t +
H(j)) (4) | |
where
H(j) =
2
{H(j)} +
2
{H(j)} |
H(j)
|
= arctan
{H(j)}
{H(j)}
Equation (4) states the answer we seek. It shows that
The steady-state sinusoidal response is a sinusoid of the same angular frequency
as the input,
The response diers from the input by (i) a change in amplitude as dened by
|H(j)|, and (ii) an added phase shift
H(j).
H(j) is known as the frequency response function. H(j) is the magnitude of the frequency | |
response function, and
H(j) is the phase.

304
mass
m
F(t)
fri cti on B
v (t)
m
| |
Note that if H(j) > 1 the sinusoidal input is amplied, while if H(j) < 1 the input is | | | |
attenuated by the system.
Example 1
The mechanical system
has a transfer function
v
m
(s) 1
H(s) = =
F (s) ms + B
where m = 1 kg, and B = 2 Ns/m. Find the steady-state response if F (t) =
10 sin(5t).
1
H(s) =
s + 2
so that the frequency response function is
H(j) = H(s)
s=j
=
1
=
2 j
|
j + 2
2
+ 4
Then
1
H(j) = , H(j) = arctan

. | |
2
+ 4

2
With = 5 rad/s,
v
ss
(t) = 10
10
|H(j)| sin(5t +
H(j)
=
29
sin(5t arctan 2.5)
= 1.857 sin(5t 1.1903)
Example 2
Plot the variation of H(j) and H(j) from = 0 to 10 rad/s.
From above
1
H(j) = , and H(j) = arctan

. | |
2
+ 4

2
These functions are plotted below:
305
frequency (rad/s)
F
r
e
q
u
e
n
c
y
R
e
s
p
o
n
s
e
M
a
g
n
i
t
u
d
e
F
r
e
q
u
e
n
c
y
R
e
s
p
o
n
s
e
P
h
a
s
e
(
d
e
g
)
0 2 4 6 8 1 0
0
-1 0
-20
-30
-40
-50
-60
-70
-80
-90
0. 5
0. 4
0. 3
0. 2
0. 1
0
Note that
As the input frequency increases, the response magnitude decreases.
At low frequencies the phase is a small negative number, but as the frequency
increases the phase lag increases and apparently is tending toward 90

at
high frequencies.
306
Appendix: Evaluation of H(j).
We start with
b
m
(j)
m
+ b
m1
(j)
m1
+ . . . + b
1
(j) + b
0
H(j) =
a
n
(j)
n
+ a
n1
(j)
n1
+ . . . + a
1
(j) + a
0
so that
b
m
(j)
m
+ b
m1
(j)
m1
+ . . . + b
1
(j) + b
0
H(j) =
a
n
(j)
n
+ a
n1
(j)
n1
+ . . . + a
1
(j) + a
0
Note that
(j)
k
= (1)
k/2
k
k even
j(1)
(k1)/2
k
k odd
(j)
k
= (1)
k/2
k
k even
j(1)
(k1)/2
k
k odd
Thus in both H(j) and H(j)
The terms with even powers of j in the numerator and denominator of H(j) and
H(j) generate real terms, while
the terms with odd powers of j generate imaginary terms.
With these substitutions, comparison of H(j) and H(j) shows
The real terms (even powers of j) are the same, while
The imaginary terms (odd powers of j) have opposite signs
leading to the conclusion
H(j) = H(j).
307
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
H(s)
u(t) = A sin (w t + q)
y (t) = A| H(jw)| si n(w t + q + H(jw))
ss
Li near System
| H(s)|
jM
I
| H(jM)| i s a "sl i ce" of the
| H(s)| surface al ong the
i magi nary axi s of the s-pl ane.

Spring Term 2008
Lecture 31
1
Reading:
Nise: 10.1
Class Handout: Frequency Response and Bode Plots
1 Sinusoidal Frequency Response (continued)
In Lecture 30 we saw that the steady-state response of a linear system with transfer function
H(s) to a sinusoidal input
u(t) = A sin(t + )
is
y
ss
(t) = A H(j) sin (t + +
H(j)) | |
where
H(j) = {H(j)} +
2
{H(j)} | |
2
H(j) = arctan
{H(j)}

{H(j)}
We note that
H(j) = H(s)
s=j
, |
that is H(j) is H(s) evaluated along the imaginary axis of the s-plane.
1
311

1.1 The Frequency Response of Systems with Zeros
If a system has a transfer function
N(s)
H(s) =
D(s)
the frequency response function is
N(j)
H(j) = .
D(j)
For complex a and b, a/b = a / b and (a/b) = b, so that | | | | | |

a
N(j)
2
{N(j)} +
2
{N(j)}
(1) |H(j)| =
|
|D(j)|
|
=
2
{D(j)} +
2
{D(j)}
H(j) = N(j) D(j) = arctan
{N(j)}

arctan
{D(j)}

(2)
{N(j)} {D(j)}
Example 1
Find and plot the frequency response of
s + 5
H(s) =
s + 10
The frequency response function is
j + 5
H(j) =
j + 10
and
N(j)
2
+ 25
|H(j)| =
|
|D(j)|
|
=
2
+ 100
H(j) = N(j) D(j) = arctan

arctan

5 10
The following MATLAB commands were used to plot the frequency response:
w=0:.2:100;
sys=zpk(-5,-10,1)
y=freqresp(sys,w);
plot(w,squeeze(abs(y)))
plot(w,squeeze(angle(y)))
which produced the following plots:
312
0 20 40 60 80 100
0 20 40 60 80 100
0
0. 2
0. 4
0. 6
0. 8
1. 0
20
18
16
14
12
10
8
6
4
2
0
F
r
e
q
u
e
n
c
y
R
e
s
p
o
n
s
e
P
h
a
s
e
(
d
e
g
)
F
r
e
q
u
e
n
c
y
R
e
s
p
o
n
s
e
M
a
g
n
i
t
u
d
e
frequency (rad/s)
frequency (rad/s)
We note that
When = 0, |H(j)| = 0.5 and
H(j) = 0
.
When , |H(j)| 1 and
H(j) 0
.
Example 2
Find the frequency response functions for (i) a dierentiator, and (ii) an integra
tor.
313
(i) A dierentiator. The transfer function is H(s) = s, so that H(j) = j.
Then

|H(j)| = ,
H(j) =
2
(or +90
)
(ii) An integrator. The transfer function is H(s) = 1/s, so that H(j) = 1/j.
Then
1
|H(j)| =
H(j) =
2
(or -90
)
1.2 The Frequency Response of a Second-Order System
Consider the unity-gain second-order system:
2
H(s) =
n
s
2
+ 2
n
s +
2
n
The frequency response is
2
H(j) =
n
2
+ j2
n
+
2
n
so that
2
(3) |H(j)| =
(
n
2

2
)
2
n
+ (2
n
)
2
2
n

H(j) = tan
1
2
(4)
n

2
Equations (3) and (4) show the following:
When = 0, H(j) = 1 | |
H(j) = 0
When H(j) 0 , | |
H(j) (or -180
)
1
When =
n
, H(j) = | |
2

H(j) =
2
(or -90
)
The response of the system to frequencies close to the undamped natural frequency clearly
depends on the damping ratio . For a lightly damped system < 0.5, H(j
n
) > 1 and | |
the system demonstrates amplication due to resonance.
Dierentiation of Eq. (4) shows the
peak
, the frequency of the peak response is not
n
but is in fact
p
=
n
1 2
2
314
0 1 2 3 4
0
1
2
3
4
5
6
0 1 2 3 4
-1 80
-1 50
-1 20
-90
-60
-30
0
Normal i zed frequency
H(jw)
| H(jw) |
z = 0.1
z = 0.1
z = 0.1
0.5
1
2
5
w/w
n
w/w
n
1
2
2
1
0. 5
5
0. 2
0. 5
0. 2
0.2
for < 1/
2, and
1
|H(j
p
)| =
2
1
2
.
Frequency response plots for several values of are shown below:
Example 3
A tall slender structure, excited by wind forces, is modeled as a mass-spring
damper system. Find the frequency response of the displacement x of the building
to a sinusoidal wind loading.
315
0 1 0 20 30 40 50
0
0. 01
0. 02
0. 03
0. 04
0. 05
0. 06
0. 07
-1 80
-1 60
-1 40
-1 20
-1 00
-80
-60
-40
-20
0
frequency (rad/s)
F
r
e
q
u
e
n
c
y
R
e
s
p
o
n
s
e
M
a
g
n
i
t
u
d
e
F
r
e
q
u
e
n
c
y
R
e
s
p
o
n
s
e
P
h
a
s
e
(
d
e
g
)
m
K
B
F
s
(wind force)
m
1
F
s
(wind force)
x
B - equi val ent dampi ng
coeffi ci ent
K - equi val ent l ateral
sti ffness
x
X(s) 1 1
H(s) = = H(j) =
F
s
(s) ms
2
+ Bs + K

(K m
2
) + jB
1
|H(j)| =
(K m
2
)
2
+ (B)
2
B
H(j) = tan
1
K m
2
With values m = 5.11 kg, B = 0.77 N-s/m, and K = 2020 N/m,
n
=
K/m = 19.9 rad/s

= B/(2m
n
) = 0.0038 N-s/m
and the system is very lightly damped. The frequency response is plotted below:
316
0
s-pl ane
p
1
(jw- p )
p
2
0
s-pl ane
p
1
p
2
z
1
q
q
1
2
f
1
q
1
2
1
q
r
s
jw
1
2
|
j
w
|
2
1
1
jw
s
jw
1
f
r
2

Note the extremely sharp and high resonant peak in the magnitude plot, and the
rapid phase transition about resonance in the phase plot.
1.3 Frequency Response and the Pole-Zero Plot
The frequency response may be written in terms of the system poles and zeros by substituting
directly into the factored form of the transfer function:
H(j) = K
(j z
1
)(j z
2
) . . . (j z
m1
)(j z
m
)
. (5)
(j p
1
)(j p
2
) . . . (j p
n1
)(j p
n
)
Because the frequency response is the transfer function evaluated on the imaginary axis of
the s-plane, that is when H(s)
j
, the graphical method for evaluating the transfer function
may be applied directly to the frequency response. Each of the vectors from the n system
poles to a test point s = j has a magnitude and an angle:
|j p
i
| =
2
+ (
i
)
2
,
i
(s p
i
) = tan
1
i
i

,
as shown above, with similar expressions for the vectors from the m zeros. The magnitude
and phase angle of the complete frequency response may then be written in terms of the
magnitudes and angles of these component vectors
m
(j z
i
)
H(j) = K
i=1
| |
m
i
n
=1
|
|
(j p
i
)
|
|
n
H(j) =

(j z
i
)

(j p
i
).
i=1 i=1
If the vector from the pole p
i
to the point s = j has length q
i
and an angle
i
from the
horizontal, and the vector from the zero z
i
to the point j has a length r
i
and an angle
i
,
the value of the frequency response at the point j is
r
1
. . . r
m
|H(j)| = K
q
1
. . . q
n
H(j) = (
1
+ . . . +
m
) (
1
+ . . . +
n
)
317
0
s-plane
s
jw
1
jw
jw
2
p = -1/t
w w
w
w
1 2
H(jw)
0
-90
1
2
| H(jw)|
K
K/q
K/q
0
1
2
q
q
1
2
q
q
2
1
(a) (b)
-q
-q
Example 4
Explain the nature of the sinusoidal response of a rst-order system with a pole
on the real axis at s = 1/ as shown below, in terms of the pole-zero plot.
Even though the gain constant K cannot be determined from the pole-zero plot,
the following observations may be made directly by noting the behavior of the
magnitude and angle of the vector from the pole to the imaginary axis as the
input frequency is varied:
At low frequencies the gain approaches a nite value, and the phase angle
has a small but nite lag.
As the input frequency is increased the gain decreases (because the length
of the vector increases), and the phase lag also increases (the angle of the
vector becomes larger).
At very high input frequencies the gain approaches zero, and the phase angle
approaches /2.
1.3.1 High Frequency Response
As we note the following
Magnitude Response: The magnitude response for the s-plane is
r
1
. . . r
m
|H(j)| = K
q
1
. . . q
n
and at high frequencies all vectors have approximately the same length, that is
r
i
q
j
for i = 1 . . . m, j = 1 . . . n
318
jw
s
x
x
x o
As w becomes l arge:
1 ) Al l vectors have approx.
the same l ength ( @ w)
2) Al l angl es are approx.
p/2
w
@ p/2
@
w
Then
1
lim H(j) = K
nm

| |
Phase Response: From the pole-zero plot
H(j) = (
1
+ . . . +
m
) (
1
+ . . . +
n
)
As becomes large all of the angles of the vectors approach /2,

i

j

2
for i = 1 . . . m, j = 1 . . . n
and

lim H(j) = (n m)

2
Then
n > m n = m n < m
lim

|H(j)|
lim

H(j)
0 K
0
(n m)/2 (m n)/2
If a system has an excess of poles over the number of zeros (n > m) the magnitude of the
frequency response tends to zero as the frequency becomes large. Similarly, if a system has
an excess of zeros the gain increases without bound as the frequency of the input increases.
(This cannot happen in physical energetic systems because it implies an innite power gain
through the system.)
319
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
0
s-pl ane
p
1
(jw- p )
p
2
0
s-pl ane
p
1
p
2
z
1
q
q
1
2
f
1
q
1
2
1
q
r
s
jw
1
2
|
j
w
|
2
1
1
jw
s
jw
1
f
r
2

Spring Term 2008
Lecture 32
1
Reading:
Nise: 10.1
Class Handout: Sinusoidal Frequency Response
1 Frequency Response and the Pole-Zero Plot (continued)
We showed that if
H(j) = K
(j z
1
)(j z
2
) . . . (j z
m1
)(j z
m
)
. (1)
(j p
1
)(j p
2
) . . . (j p
n1
)(j p
n
)
and if each of the vectors from the n system poles to a test point s = j has a magnitude
and an angle:
|j p
i
| = =
2
+ (
i
)
2
, q
i
i
(s p
i
) =
i
= tan
1
i
i

,
and similarly for the m zeros
|j z
i
| = r
i
=
2
+ (
i
)
2
,
i
(s z
i
) =
i
= tan
1
i
i

,
the value of the frequency response at the point j is
r
1
. . . r
m
|H(j)| = K
q
1
. . . q
n
H(j) = (
1
+ . . . +
m
) (
1
+ . . . +
n
)
1
321
0
s-pl ane
p
1
p
2
z
1
G
G
1
2
B
1
q
1
2
1
q
r
jM
I
M B
r
2
2
1.0.1 High Frequency Response
In Lecture 31 we saw that at high frequencies all vectors have approximately the same length,
that is and
1
lim H(j) = K
nm

| |
and that all of the angles of the vectors approach /2, with the result

lim
H(j) = (n m)
2

If a system has an excess of poles over the number of zeros (n > m) the magnitude of the
frequency response tends to zero as the frequency becomes large. Similarly, if a system has
an excess of zeros the gain increases without bound as the frequency of the input increases.
If n = m the magnitude function tends to a constant K.
1.0.2 Low Frequency Response
As 0 we note the following
Magnitude Response: The magnitude response for the s-plane is
r
1
. . . r
m
|H(j)| = K
q
1
. . . q
n
If any of the r
i
0, then H(j) 0, and if any q
i
0, then H(j) | | | |
If a system has one or more zeros at the origin of the s-plane (corresponding to a pure
dierentiation), then the system will have zero gain at = 0. Similarly, if the system has
one or more poles at the origin (corresponding to a pure integration term in the transfer
function), the system has innite gain at zero frequency.
lim
0
|H(j)| = 0 if there are zeros at the origin
lim H(j) if there are poles at the origin

0
| | =
r
1
. . . r
m
lim
0
|H(j)| = K
q
1
. . . q
n
otherwise
322
0
I
jM
1
jM
p
p
1
2
1
q
M i ncreasi ng
i n the proxi mi ty of p the
l ength q decreases to a
mi ni mum.
1
1
Phase Response:
H(j) = (
1
+ . . . +
m
) (
1
+ . . . +
n
)
As 0:
All real-axis l.h.p. poles and zeros contribute 0 rad. to the phase response.
Each complex conjugate pole or zero pair contributes a total of 2 rad. to the phase
response (eectively adding 0 rad. to the total response).
A pole at the origin (s = 0 + j0) contributes /2 rad. to the phase response.
A zero at the origin (s = 0 + j0) contributes +/2 rad. to the phase response.
A r.h.p real zero contributes + rad. to the phase response.
The low frequency phase response is therefore

lim H(j) = (N M) + L rad.
0
where N is the number of poles at the origin, M is the number of zeros at the origin, and L
is the number of r.h.p. real zeros.
1.0.3 Behavior in the Proximity of Poles and Zeros Close to the Imaginary Axis
Consider a second-order system with a damping ratio 1, so that the pair of complex
conjugate poles are located close to the imaginary axis.
K
|H(j)| =
q
1
q
2
H(j) = (
1
+
2
)
In this case there are a pair of vectors connecting the two poles to the imaginary axis,
and the following conclusions may be drawn by noting how the lengths and angles of the
vectors change as the test frequency moves up the imaginary axis: As the input frequency
is increased and the test point on the imaginary axis approaches the pole, one of the vectors
(associated with the pole in the second quadrant) decreases in length, and at some point
reaches a minimum.
323
0
s
jw
1
jw
jw
2
H(jw)
w
w
| H(jw)|
0
0
-180
w
1
w
p
p
1
2
(a) (b)
Because q
1
appears in the denominator of the magnitude function, over this range there is
an increase in the value of H(j) . | |
If a system has a pair of complex conjugate poles close to the imaginary axis, the
magnitude of the frequency response has a peak, or resonance at frequencies in the
proximity of the pole. If the pole pair lies directly upon the imaginary axis, the system
exhibits an innite gain at that frequency.
Similarly, if a system has a pair of complex conjugate zeros close to the imaginary
axis, the frequency response. Over this range has a dip or notch in its magnitude
function at frequencies in the vicinity of the zero. Should the pair of zeros lie directly
upon the imaginary axis, the response is identically zero at the frequency of the zero,
and the system does not respond at all to sinusoidal excitation at that frequency.
Similarly in the proximity of the pole there is a rapid change of the angle
1
associated with
the pole p
1
.
2 Logarithmic (Bode) Plots
In system dynamic analyses, frequency response characteristics are almost always plotted
using logarithmic scales. In particular, the magnitude function H(j) is plotted against | |
frequency on a log-log scale, and the phase
H(j) is plotted on a linear-log scale. For

example, the frequency response functions of a typical rst-order system dy/dt + y = u(t)
is plotted below on (a) linear axes, and (b) logarithmically scaled axes.
324
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-90
-80
-70
-60
-50
-40
-30
-20
-1 0
0
1 0
wt
0 2 4 6 8 1 0
-90
-75
-60
-45
-30
-1 5
0
1 5
0 2 4 6 8 1 0
0. 0
0. 2
0. 4
0. 6
0. 8
1 . 0
1 . 2
| H(jw)|
H(jw)
wt
wt
1 . 0
0. 1
0. 01
| H(jw)|
H(jw)
(b) (a)
wt
P
h
a
s
e
r
e
s
p
o
n
s
e
(
d
e
g
)
N
o
r
m
a
l
i
z
e
d
m
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
N
o
r
m
a
l
i
z
e
d
m
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
P
h
a
s
e
r
e
s
p
o
n
s
e
(
d
e
g
)
Similarly the second-order frequency response is shown in linear and logarithmic forms below
H(jw)
| H(jw)|
z = 0.1
z = 0.1
z = 0.1
0.5
1
2
5
w/w
n
w/w
n
1
2
2
1
0. 5
5
0. 2
0. 5
0. 2
0.2
0
1 2 3 4
-1 80
-1 50
-1 20
-90
-60
-30
0
0 1 2 3 4
0
1
2
3
4
5
6
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-1 80
-1 60
-1 40
-1 20
-1 00
-80
-60
-40
-20
0
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
Normal i zed angul ar frequency
H(jw)
z=5
z = 0.1
1
z = 0.1
0.2
0.5
2
z=5
1
z=5
z = 0.1
0.2
2
0.5
(a)
1
1 0
. 1
. 01
. 001
. 0001
| H(jw)|
w/w
n
w/w
n
(b)
P
h
a
s
e
r
e
s
p
o
n
s
e
(
d
e
g
)
P
h
a
s
e
r
e
s
p
o
n
s
e
(
d
e
g
)
N
o
r
m
a
l
i
z
e
d
m
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
N
o
r
m
a
l
i
z
e
d
m
a
g
n
i
t
u
d
e
r
e
s
p
o
n
s
e
325
It can be seen that while two sets of plots convey the same information, they have a dierent
appearance. The logarithmic frequency scale has the eect of expanding the low frequency
region of the plots while compressing the high frequencies. The logarithmic magnitude plot
can be seen to exhibit straight line asymptotic behavior at high and low frequencies.
In the 1940s H. W. Bode introduced the logarithmic frequency response plots as a sim
plied method for sketching approximate frequency response characteristics of electronic
feedback ampliers. Bode plots, named after him, have subsequently been widely used in
linear system design and analysis, and in feedback control system design and analysis. The
Bode sketching method provides an eective means of approximating the frequency response
of a complex system by combining of the responses of simple rst and second-order systems.
2.1 Logarithmic Amplitude and Frequency Scales:
2.1.1 Logarithmic Amplitude Scale: The Decibel
Bode magnitude plots are frequently plotted using the decibel logarithmic scale to display the
function H(j) . The Bel, named after Alexander Graham Bell, is dened as the logarithm | |
to base 10 of the ratio of two power levels. In practice the Bel is too large a unit, and the
decibel (abbreviated dB), dened to be one tenth of a Bel, has become the standard unit of
logarithmic power ratio. The power ow P into any element in a system, may be expressed
in terms of a logarithmic ratio Q to a reference power level P
ref
:
Q = log
10

P

Bel or Q = 10 log
10

P

dB. (2)
P
ref
P
ref
Because the power dissipated in a Dtype element is proportional to the square of the
amplitude of a system variable applied to it, when the ratio of across or through variables is
computed the denition becomes

A
2

A
Q = 10 log
10
= 20 log
10
dB. (3)
A
ref
A
ref
where A and A
ref
are amplitudes of variables.
Note: This denition is only strictly correct when the two amplitude quantities are
measured across a common Dtype (dissipative) element. Through common usage,
however, the decibel has been eectively redened to be simply a convenient loga
rithmic measure of amplitude ratio of any two variables. This practice is widespread
in texts and references on system dynamics and control system theory.
The table below expresses some commonly used decibel values in terms of the power and
amplitude ratios.
326
Decibels Power Ratio Amplitude Ratio
-40 0.0001 0.01
-20 0.01 0.1
-10 0.1 0.3162
-6 0.25 0.5
-3 0.5 0.7071
0 1.0 1.0
3 2.0 1.414
6 4.0 2.0
10 10.0 3.162
20 100.0 10.0
40 10000.0 100.0
The magnitude of the frequency response function H (j) is dened as the ratio of the | |
amplitude of a sinusoidal output variable to the amplitude of a sinusoidal input variable.
This ratio is expressed in decibels, that is
Y (j)
20 log
10
|H(j)| = 20 log
10
|
U(j)
|
dB.
| |
As noted this usage is not strictly correct because the frequency response function does not
dene a power ratio, and the decibel is a dimensionless unit whereas H (j) may have | |
physical units.
Example 1
An amplier has a gain of 28. Express this gain in decibels.
We note that 28 = 10 2 1.4 10 2
2. The gain in dB is therefore

20 log
10
10 + 20 log
10
2 + 20 log
10
2, or

Gain(dB) = 20 + 6 + 3 = 29 dB.
The advantages of a logarithmic amplitude scale include:
Compression of a large dynamic range.
Cascaded subsections may be handled by addition instead of multiplication, that is
log( H
1
(j)H
2
(j)H
3
(j) ) = log( H
1
(j) ) + log( H
2
(j) ) + log( H
3
(j) ) | | | | | | | |
which is the basis for the sketching rules.
High and low frequency asymptotes become straight lines when log( H(j) ) is plotted
against log().
| |
327
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008

Spring Term 2008
Lecture 33
1
Reading:
Nise: 10.1
1 Bode Plots (continued
1.1 Logarithmic Amplitude and Frequency Scales:
1.1.1 Logarithmic Amplitude Scale: The Decibel
Bode magnitude plots are frequently plotted using the decibel logarithmic scale to display the
function H(j) . The Bel, named after Alexander Graham Bell, is dened as the logarithm | |
to base 10 of the ratio of two power levels. In practice the Bel is too large a unit, and the
decibel (abbreviated dB), dened to be one tenth of a Bel, has become the standard unit of
logarithmic power ratio. The power ow P into any element in a system, may be expressed
in terms of a logarithmic ratio Q to a reference power level P
ref
:
Q = log
10

P

Bel or Q = 10 log
10

P

dB. (1)
P
ref
P
ref
Because the power dissipated in a Dtype element is proportional to the square of the
amplitude of a system variable applied to it, when the ratio of across or through variables is
computed the denition becomes

A
2

A
Q = 10 log
10
= 20 log
10
dB. (2)
A
ref
A
ref
where A and A
ref
are amplitudes of variables.
Note: This denition is only strictly correct when the two amplitude quantities are
measured across a common Dtype (dissipative) element. Through common usage,
however, the decibel has been eectively redened to be simply a convenient loga
rithmic measure of amplitude ratio of any two variables. This practice is widespread
in texts and references on system dynamics and control system theory.
The table below expresses some commonly used decibel values in terms of the power and
amplitude ratios.
1
331
Decibels Power Ratio Amplitude Ratio
-40 0.0001 0.01
-20 0.01 0.1
-10 0.1 0.3162
-6 0.25 0.5
-3 0.5 0.7071
0 1.0 1.0
3 2.0 1.414
6 4.0 2.0
10 10.0 3.162
20 100.0 10.0
40 10000.0 100.0
The magnitude of the frequency response function H (j) is dened as the ratio of the | |
amplitude of a sinusoidal output variable to the amplitude of a sinusoidal input variable.
This ratio is expressed in decibels, that is
Y (j)
20 log
10
|H(j)| = 20 log
10
|
U(j)
|
dB.
| |
As noted this usage is not strictly correct because the frequency response function does not
dene a power ratio, and the decibel is a dimensionless unit whereas H (j) may have | |
physical units.
Example 1
An amplier has a gain of 28. Express this gain in decibels.
We note that 28 = 10 2 1.4 10 2
2. The gain in dB is therefore

20 log
10
10 + 20 log
10
2 + 20 log
10
2, or

Gain(dB) = 20 + 6 + 3 = 29 dB.
The advantages of a logarithmic amplitude scale include:
Compression of a large dynamic range.
Cascaded subsections may be handled by addition instead of multiplication, that is
log( H
1
(j)H
2
(j)H
3
(j) ) = log( H
1
(j) ) + log( H
2
(j) ) + log( H
3
(j) ) | | | | | | | |
which is the basis for the sketching rules.
High and low frequency asymptotes become straight lines when log( H(j) ) is plotted
against log().
| |
332
-1 0 1 2 3
0. 1 1 1 0
1 00
1 000 2 5
l og M
M
10
1 decade
1 octave
Note: M = 0 cannot be shown
1.1.2 Logarithmic Frequency Scales
In the Bode plots the frequency axis is plotted on a logarithmic scale. Two logarithmic units
of frequency ratio are commonly used: the octave which is dened to be a frequency ratio of
2:1, and the decade which is a ratio of 10:1.
Given two frequencies
1
and
2
the frequency ratio W = (
1
/
2
) between them may be
expressed logarithmically in units of decades or octaves by the relationships
W = log
2
(
1
/
2
) octaves
= log
10
(
1
/
2
) decades.
The terms above and below are commonly used to express the positive and negative
values of logarithmic values of W . A frequency of 100 rad/s is said to be two octaves (a
factor of 2
2
) above 25 rad/s, while it is three decades (a factor of 10
3
) below 100,000 rad/s.
1.2 Asymptotic Bode Plots of Low-Order Transfer Functions
The Bode plots consist of (1) a plot of the logarithmic magnitude (gain) function, and (2) a
separate linear plot of the phase shift, both plotted on a logarithmic frequency scale. In this
section we develop the plots for rst and second-order terms in the transfer function. The
approximate sketching methods described here are based on the fact that an approximate
loglog magnitude plot can be derived from a set of simple straight line asymptotic plots
that can be easily combined graphically.
The system transfer function in terms of factored numerator and denominator polyno
mials is
H(s) = K
(s z
1
)(s z
2
) . . . (s z
m1
)(s z
m
)
(s p
1
)(s p
2
) . . . (s p
n1
)(s p
n
)
, (3)
where the z
i
, for i = 1, . . . , m, are the system zeros, and the p
i
, for i = 1, . . . , n, are the
system poles.
In general a system may have complex conjugate pole and zero pairs, real poles and zeros,
and possibly poles or zeros at the origin of the s-plane. Bode plots are constructed from a
rearranged form of Eq. (??), in which complex conjugate poles and zeros are combined into
second-order terms with real coecients. For example a pair of complex conjugate poles
s
i
, s
i+1
=
i
j
i
is written
1

1

1
(s (
i
+ j
i
)) (s (
i
j
i
))
=
2
(1 (/
n
)
2
) + j2/
n
(4)
s=j n
333
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-40
-30
-20
-1 0
0
1 0
20
30
40
Angul ar frequency (rad/sec)
M
20l og | H(jM)|
1 0
sl ope
-20dB/decade
| |
and described by parameters

n
and . The constant terms 1/
n
2
is absorbed into a rede
nition of the gain constant K.
In the following sections Bode plots are developed for the rst and second-order numerator
and denominator terms:
1.2.1 Constant Gain Term:
The simplest transfer function is a constant gain, that is H(s) = K.
|H(j)| = K and
H(j) = 0,
and converting to the logarithmic decibel scale
20 log
10
H(j) = 20 log
10
K and H(j) = 0 dB.
The Bode magnitude plot is a horizontal line at the appropriate gain and the phase plot is
identically zero for all frequencies.
1.2.2 A Pole at the Origin of the s-plane:
A single pole at the origin of the s-plane, that is H(s) = 1/s, has a frequency response
1
|H(j)| =

and
H(j) = /2.
The value of the magnitude function in logarithmic units is
log |H(j)| = log()
or using the decibel scale
20 log
10
|H(j)| = 20 log
10
() dB.
The decibel based Bode magnitude plot is therefore a straight line with a slope of -20
dB/decade and passing through the 0 dB line ( H(j) = 1) at a frequency of 1 rad/s. The | |
phase plot is a constant value of /2 rad, or 90
, at all frequencies. The magnitude Bode

plot for this system is shown below.
334
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-40
-30
-20
-1 0
0
1 0
20
30
40
M
sl ope
20 dB/decade
20l og | H(jM)|
1 0
1.2.3 A Single Zero at the Origin:
A single zero at the origin of the s-plane, that is H(s) = s, has a frequency response H(j)
with magnitude and phase
|H(j)| = and
H(j) = /2.
The logarithmic magnitude function is therefore
log H(j) = log() | |
or in decibels
20 log
10
H(j) = 20 log
10
() dB. | |
The Bode magnitude plot is a straight line with a slope of +20 dB/decade. This curve also
has a gain of 0 dB (unity gain) at a frequency of 1 rad/s. The phase plot is a constant of
/2 radians, or +90
, at all frequencies. The magnitude plot is shown in below.

1.2.4 A Single Real Pole
The frequency response of a unity-gain single real pole factor is
1
H(s) = ,
s + 1
and the frequency response is:
1
|H(j)| =
()
2
+ 1
and
H(j) = tan
1
().
The logarithmic magnitude function is
log |H(j)| = 0.5 log
()
2
+ 1

,
or as a decibel function
20 log
10
|H(j)| = 10 log
10
()
2
+ 1

dB.
335
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-40
-30
-20
-1 0
0
1 0
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-90
-80
-70
-60
-50
-40
-30
-20
-1 0
0
1 0
wt
wt
H(j w)
sl ope
-20dB/decade
20l og | H(jw)|
1 0
When 1, the rst term may be ignored and the magnitude may be approximated
by a low-frequency asymptote

lim
0
20 log
10
|H(j)| = 10 log
10
(1) = 0 dB
which is a horizontal line on the plot at 0dB (unity) gain.

At high frequencies, for which 1, the unity term in the magnitude expression may
be ignored and the magnitude function is approximated by a high-frequency asymptote
20 log
10
|H(j)| 10 log
10
(( )
2
) = 20 log
10
() 20 log
10
() dB.
which is a straight line when plotted against log(), with a slope of -20 dB/decade.
The high and low frequency asymptotes intersect on the plot on the 0 dB line at a
corner or break frequency of = 1/. We note that when = 1/ the magnitude is
|H(j)| = 1/
2 or -3 dB.
The complete asymptotic Bode magnitude plot as dened by these two line segments is shown
in (a) below using a normalized frequency axis. The exact response is also shown in the gure;
at the break frequency = 1/ the actual response is 20 log
10
|H(j)| = 10 log
10
(2) = 3
dB.
336
The phase characteristic is also plotted against a normalized frequency scale in (a). At low
frequencies the phase shift approaches 0 radians. It passes through a phase shift of /4
radians at the break frequency = 1/, and asymptotically approaches a maximum phase
lag of /2 radians as the frequency becomes very large. A piece-wise linear approximation
may be made by assuming that the curve has a phase shift of 0 radians at frequencies more
than one decade below the break frequency, a phase shift of /2 radians at frequencies
more than a decade above the break frequency, and a linear transition in phase between
these two frequencies on the logarithmic frequency scale. This approximation is within 0.1
radians of the exact value at all frequencies.
1.2.5 A Single Real Zero
A numerator term, corresponding to a single real zero, written in the form H(s) = s + 1
(where is not strictly a time constant), is handled in a manner similar to a real pole. In
this case
H(j) = j + 1
and the magnitude and phase responses are
|H(j)| =
1 + ()
2
and
H(j) = tan
1
( )
respectively. In decibels the magnitude expression is
20 log
10
H(j) = 10 log
10
(1 + ()
2
) dB. | |
The low frequency asymptote is found by assuming that 1 in which case
lim 20 log
10
H(j) = 10 log
10
(1) = 0 dB,
0
| |
The high frequency asymptote is found by assuming that 1,

20 log
10
|H(j)| 20 log
10
() = 20 log
10
() 20 log
10
() dB when 1/
which is a straight line on the log-log plot, with a slope of +20 dB/decade.
The break frequency, dened by the intersection of these two asymptotes is at a fre
quency = 1/, and at this frequency the exact value of |H(j)| is
2 or +3 dB.
The complete asymptotic Bode magnitude plot using a normalized frequency scale is
shown below.
The phase characteristic asymptotically approaches 0 radians at low frequencies and ap
proaches a maximum phase lead of /2 radians at frequencies much greater than the break
frequency. At the break frequency the phase shift is /4 radians. A piece-wise linear ap
proximation, similar to that described for a real pole, is also shown below.
337
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-1 0
0
1 0
20
30
40
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-1 0
0
1 0
20
30
40
50
60
70
80
90
wt
wt
H(j w)
sl ope
20 dB/decade
20l og | H(jw)|
1 0
1.2.6 Complex Conjugate Pole Pair:
The classical second-order system,
H(s) =
2
n
s
2
+ 2
n
s +
n
2
has a frequency response
1
2 2
|H(j)| =
(1 (/
n
)
2
) + (2(/
n
))
and H(j) = tan
1
2 (/
n
)
1 (/
n
)
2

.
In logarithmic units the magnitude response is
)
2
2
20 log
10
|H(j)| = 10 log
10
1 (/
n
+ (2(/
n
))
2

The Bode forms of the magnitude and phase responses are plotted in below, with the damping
ratio as a parameter.
338
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-80
-60
-40
-20
0
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-80
-60
-40
-20
0
20
20l og | H(jw)|
1 0
H(jw)
w/w
n
w/w
n
z=5
z = 0.1
1
z = 0.1
0.2
0.5
2
z=5
1
z=5
z = 0.1
0.2
2
-40dB/decade
sl ope
0.5
The low-frequency asymptote is found by assuming that /
n
1 so that
lim (20 log
10
H(j) ) = 10 log
10
(1) = 0 dB.
(/
n
) 0
| |
The high frequency response can be found by retaining only the dominant term when
/
n
1:
20 log
10
|H(j)| 10 log
10
(/
n
)
4

= 40 log
10
() + 40 log
10
(
n
) dB when
n
,
which is a linear function of log
10
with a slope of -40 dB/decade.
The two asymptotes intersect at a break frequency of =
n
as shown below. The
straight line asymptotic form does not account in any way for the damping ratio.
The phase characteristic asymptotically approaches 0 radians at low frequencies, has a
phase lag of /2 at the break frequency
n
, and approaches radians at high frequencies.
The steepness of the transition is a function of the damping ratio and so must be sketched
using the information contained above.
The resonance peak (for values of < 0.707) must be sketched in after the asymptotes
have been drawn. The gure below plots the logarithmic magnitude correction and frequency
of the resonant peak as a function of ; it is a simple matter to sketch in the resonant peak
from these values.
339
0 0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7
0
1 0
20
30
40
0 0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7
0
0. 2
0. 4
0. 6
0. 8
1
1 . 2
Dampi ng rati o
z
Dampi ng rati o
z
w /w
p n
1 0
20l og | H(jw )|
p
(a) (b)
1.2.7 Complex Conjugate Zero Pair
Bode plots for a pair of complex conjugate zeros can be derived in a manner similar to the
conjugate pole pair described above. In this case the block is assumed to have a transfer
function
H(s) =
1
s
2
+ 2
n
s +
2
n

n
and a frequency response
2 2
|H(j)| = (1 (/
n
)
2
) + (2(/
n
))
and H(j) = tan
1
2 (/
n
)
1 (/
n
)
2
.
The logarithmic magnitude response is
)
2
2
20 log
10
|H(j)| = 10 log
10
1 (/
n
+ (2(/
n
))
2

dB
The asymptotic responses are derived in a similar manner to the complex pole pair; the low
frequency asymptote is
lim (20 log
10
H(j) ) = 10 log
10
(1) = 0 dB,
(/
n
) 0
| |
and the high frequency asymptote is

20 log
10
|H(j)| 10 log
10
(/
n
)
4

= 40 log
10
() 40 log
10
(
n
) dB for
n
.
The exact form of the magnitude response is plotted below. This is eectively an inverse
of the characteristic of complex-conjugate pole pair described above. There is a notch in
the response in the region of the frequency
n
, and the depth is a function of the parameter
. The plot has a low frequency asymptote of 0 dB, a break frequency of =
n
, and a
3310
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-20
0
20
40
60
80
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
0
20
40
60
80
1 00
1 20
1 40
1 60
1 80
H(jw)
20l og | H(jw)|
1 0
w/w
n
w/w
n
z=5
z=0. 1
0. 2
0. 5
2
1
z = 0. 1
0. 2
0. 5
1
2
z=5
z = 0. 1
z=5
40dB/decade
sl ope
high-frequency asymptote is a straight line with a slope of +40 dB/decade. The phase char
acteristic is also a ipped version of that of a pair of complex conjugate poles; it approaches
0 radians at low frequencies, passes through /2 at the break frequency, and shows a max
imum phase lead of radians at high frequencies. As above, the slope of the curve in the
transition region is dependent on the value of .
1.2.8 Summary
The essential features of the asymptotic forms of the seven components of the magnitude
plot are summarized below.
3311
Description Transfer Function Break Frequency
(radians/sec.)
High Frequency Slope
(dB/decade)
Constant gain
Pole at the origin
Zero at the origin
Real pole
Real zero
Conjugate poles
Conjugate zeros
K
1
s
s
1
s + 1
(s + 1)
2
n
s
2
+ 2
n
s +
2
n
1
2
n
(s
2
+ 2
n
s +
2
n
)
-
-
-
1/
1/
n
0
-20
+20
-20
+20
-40
+40
3312
MIT OpenCourseWare
http://ocw.mit.edu
Spring 2008
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-40
-30
-20
-1 0
0
1 0
20
30
40
M
20l og | H(jM)|
1 0
sl ope
-20dB/decade
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-40
-30
-20
-1 0
0
1 0
20
30
40
M
sl ope
20 dB/decade
(a) (b)
20l og | H(jM)|
1 0

Spring Term 2008
Lecture 34
1
Reading:
Nise: 10.1
1 Bode Plots (continued
In Lecture 33 we developed the following asymptotic Bode Plot s for low-order systems:
Pole/Zero at the Origin:
1
H(s) = and H(s) = s
s
1
341
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-40
-30
-20
-1 0
0
1 0
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-90
-80
-70
-60
-50
-40
-30
-20
-1 0
0
1 0
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-1 0
0
1 0
20
30
40
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-1 0
0
1 0
20
30
40
50
60
70
80
90
wt wt
wt
wt
Normal i zed angul ar frequency Normal i zed angul ar frequency
H(j w) H(j w)
sl ope
-20dB/decade
sl ope
20 dB/decade
20l og | H(jw)|
1 0
20l og | H(jw)|
1 0
Single Real Pole/Zero:
1
H(s) = and H(s) = s + 1
s + 1
342
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-1 80
-1 60
-1 40
-1 20
-1 00
-80
-60
-40
-20
0
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-80
-60
-40
-20
0
20
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-20
0
20
40
60
80
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
0
20
40
60
80
1 00
1 20
1 40
1 60
1 80
H(jw)
20l og | H(jw)|
1 0
20l og | H(jw)|
1 0
H(jw)
w/w
n
w/w
n
w/w
n
w/w
n
z=5
z=0. 1
0. 2
0. 5
2
1
z = 0. 1
0. 2
0. 5
1
2
z=5
z = 0. 1
z=5 z=5
z = 0. 1
1
z = 0. 1
0. 2
0. 5
2
z=5
1
z=5
z = 0. 1
0. 2
2
40dB/decade
sl ope
-40dB/decade
sl ope
0. 5
(a) (b)
Complex Conjugate Real Pole/Zero pair:
2
1
n
H(s) =
s
2
+ 2
n
s +
n
2
and H(s) =
n
2
s
2
+ 2
n
s +
n
2

Description Transfer Function Break Frequency High Frequency Slope
(radians/sec.) (dB/decade)
Constant gain K - 0
1
Pole at the origin
Zero at the origin
Real pole
Real zero
Conjugate poles
Conjugate zeros
s
s
1
s + 1
(s + 1)
2
n
s
2
+ 2
n
s +
2
n
1
2
n
(s
2
+ 2
n
s +
2
n
)
-
-
1/
1/
n
-20
+20
-20
+20
-40
+40
343

1.1 Bode Plots of Higher Order Systems

If a system with transfer function H(s) = KN(s)/D(s) is expressed as a product of the
terms in the table above, that is
N(s)
H(s) = K
D(s)
1 1
= K
(N
1
(s) . . . N
m
(s)) (
D
1
(s)
. . .
D
n
(s)
)
where the factors N
i
(s) are rst- or second-order zero terms, and the D
i
(s) are pole terms,
and K

is a modied constant factor. For example
10(s + 3) 1 1
H(s) = = K
N
1
(s)
(s + 0.5)(s + 5) D
1
(s) D
2
(s)
1 1 1
= 12 (
3
s + 1)
2s + 1

0.2s + 1
.
When complex numbers are represented in polar form, the magnitude of a product is the
product of the component magnitudes, and the angle of a product is the sum of the compo
nent angles, the frequency response may be expressed in terms of its magnitude and phase
functions:
|H(j)| = K

|N
1
(j)| . . . |N
m
(j)|
D
1
(j)
. . .
D
n
(j)
1 1
H(j) =
N
1
(j) + . . . +
N
m
(j) +
D
1
(j)
+ . . . +
D
n
(j)
The logarithm of a product is the sum of the logarithms of its factors, so that
1
log H(j) = log K

+ log N
1
(j) + . . . + log N
m
(j) + log + . . . + log | | | | | |
D
1
(j) D
n
(j)
1 1
H(j) =
N
1
(j) + . . . +
N
m
(j) +
+ . . . +
D
1
(j) D
n
(j)
which express the overall magnitude and phase responses as a sum of component responses of
rst and second-order elementary building blocks. In practice Bode plots are constructed
by graphically adding the individual response components. Given the transfer function H(s)
of a linear system, the following steps are used to construct the Bode magnitude plot:
1. Factor the numerator and denominator of the transfer function into the constant, rst-
order and quadratic terms in the form described in the previous section.
2. Identify the break frequency associated with each factor.
3. Plot the asymptotic form of each of the factors on loglog axes.
4. Graphically add the component asymptotic plots to form the overall plot in straight
line form.
344
-40
-20
0
20
40
60
M
a
g
n
i
t
u
d
e
(
d
B
)
1 0
-2
1 0
-1
1 0
0
1 0
1
1 0
2
1 0
3
-90
-45
0
45
90
P
h
a
s
e
(
d
e
g
)
Bode Di agram
Frequency (rad/sec)
s + 1
0. 1s + 1
1
0. 1s + 1
s + 1
0. 1s + 1
s + 1
0. 1s + 1
1
s + 1
5. Round out the corners in the straight line approximate curve by hand, using the
known values of the responses at the break frequencies (3dB for rst-order sections,
and dependent upon for quadratic factors).
The phase plot is constructed by graphically by adding the component phase responses. The
individual plots are drawn, either as the piece-wise linear approximation for the rst-order
poles, or in a smooth form from the exact plot, and then these are added to nd the total
phase shift at any frequency.
Example 1
Plot the Bode magnitude and phase plots of a rst-order system described by
the transfer function
10(s + 1)
H(s) =
s + 10
Solution: The transfer function is rewritten in terms of unity-gain blocks
1 1 1
H(s) = 10 (s + 1)
10

0.1s + 1
= (s + 1)
0.1s + 1
with two component terms:
1. A single real zero term H
1
(s) = (s + 1), with a break frequency of = 1
radians/sec.
2. A single real pole term H
2
(s) =
0.1
1
s+1
, with a break frequency of = 10
radians/sec.
The component terms are plotted and are added together to determine the total
response for a frequency range of 0.01 to 1000 radians/sec. in the magnitude and
phase plots below.
345
Example 2
Plot the Bode magnitude and phase plots of a third-order system described by
the transfer function
40s + 4
H(s) =
s
3
+ 2s
2
+ 2s
Solution: The transfer function is rewritten
4(10s + 1)
1

2
H(s) = = 2 (10s + 1)
s (s
2
+ 2s + 2) s s
2
+ 2s + 2
indicating four component terms:
1. A constant gain term of H
1
(s) = 2,
2. A single real pole at the origin H
2
(s) = 1/s,
3. A complex conjugate pole pair H(s) = 2/(s
2
+ 2s + 2), characterized by
n
=
2 radians/sec. and a damping ratio of
2/2, and
4. A single real zero term H
3
(s) = (10s +1), with a break frequency of = 0.1
radians/sec.
The component terms are plotted and are added together to determine the total
response for a frequency range of 0.01 to 100 radians/sec. in the magnitude and
phase plots below.
346
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-80
-60
-40
-20
0
20
40
60
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-1 80
-1 35
-90
-45
0
45
90
H (s) = 2
1
H (s) = (10s+1)
4
H (s) = 2/(s +2s+2)
3
2
H (s) = 1/s
2
H (s) = (10s+1)
4
H (s) = 2
1
H (s) = 2/(s +2s+2)
3
2
H (s) = 1/s
2
total response
total response
H(jw)
20l og | H(jw)|
1 0
1.2 A Simple Method for constructing the Magnitude Bode Plot
directly from the Pole-Zero Plot
The pole-zero plot of a system contains sucient information to dene the frequency re
sponse except for an arbitrary gain constant. It is often sucient to know the shape of the
magnitude Bode plot without knowing the absolute gain. The method described here allows
the magnitude plot to be sketched by inspection, without drawing the individual compo
nent curves. The method is based on the fact that the overall magnitude curve undergoes a
change in slope at each break frequency.
The rst step is to identify the break frequencies, either by factoring the transfer function
or directly from the pole-zero plot. Consider a typical pole-zero plot of a linear system as
shown in (a) in the gure below. The break frequencies for the four rst and second-order
blocks are all at a frequency equal to the radial distance of the poles or zeros from the origin
347
5 rad/sec
1 . 41 4 rad/sec
0. 1 rad/sec
I
jM
j1
-0. 1 -5 -1
-j1
s-pl ane
0. 01 0. 03 0. 1 0. 3 1 3 1 0 30 1 00
-30
-20
-1 0
0
1 0
20
30
N = 1 N = 0 N = -2 N = -1
20l og | H(jM)|
1 0
M
(a)
(b)
exact response
N
=
0
N
=
1
N
=
2
N
=
-
1
G
a
i
n
(
d
B
)
of the s-plane, that is
b
=
2
+
2
. Therefore all break frequencies may be found by
taking a compass and drawing an arc from each pole or zero to the positive imaginary axis.
These break frequencies may be transferred directly to the logarithmic frequency axis of the
Bode plot.
Because all low frequency asymptotes are horizontal lines with a gain of 0dB, a pole or
zero does not contribute to the magnitude Bode plot below its break frequency. Each pole
or zero contributes a change in the slope of the asymptotic plot of 20 dB/decade above its
break frequency. A complex conjugate pole or zero pair denes two coincident breaks of 20
dB/decade (one from each member of the pair), giving a total change in the slope of 40
dB/decade. Therefore, at any frequency , the slope of the asymptotic magnitude function
depends only on the number of break points at frequencies less than , or to the left on the
Bode plot. If there are Z breakpoints due to zeros to the left, and P breakpoints due to
poles, the slope of the curve at that frequency is 20 (Z P ) dB/decade.
Any poles or zeros at the origin cannot be plotted on the Bode plot, because they are
eectively to the left of all nite break frequencies. However, they dene the initial slope.
If an arbitrary starting frequency and an assumed gain (for example 0dB) at that frequency
are chosen, the shape of the magnitude plot may be easily constructed by noting the initial
slope, and constructing the curve from straight line segments that change in slope by units
of 20 dB/decade at the breakpoints. The arbitrary choice of the reference gain results in
a vertical displacement of the curve.
In (b) the straight line magnitude plot for the system is shown, constructed using this
method. A frequency range of 0.01 to 100 radians/sec was arbitrarily selected, and a gain
of 0dB at 0.01 radians/sec was assigned as the reference level. The break frequencies at 0,
0.1, 1.414, and 5 radians/sec were transferred to the frequency axis from the pole-zero plot.
The value of N at any frequency is Z P , where Z is the number of zeros to the left, and
P is the number of poles to the left. The curve was simply drawn by assigning the value of
the slope in each of the frequency intervals and drawing connected lines.
348

Dynamics and Control II

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Dynamics and Control II

Hochgeladen von

Copyright:

Verfügbare Formate

MIT OpenCourseWare

which has the form

F (s) = L{f(t)} = f(t)e

and L{(t)} = 1 from above.

) = 0, etc then using the dierentiation property of the Laplace transform

as (a) a series connection, and (b) a parallel connection of rst-order blocks

"l i near" regi on

(b) The spring (compliance) element:

voltage is so small (volts) that we declare it to be a virtual ground, that is in our

arrow shows di recti on

arrow shows di recti on

2) sin(t /4), that is the amplitude is

2/2 is a commonly used specication for system design

(s) is a polynomial of degree less than that of D(s).

We can also dene an acceleration constant K

The Angle Condition:

where {} is the real operator, and {} is the imaginary operator.

. The centroid will be

. The two root loci below show cases for

. The angle of departure

We then use the trig. identity

H(j) is the phase.

H(j) (or -180

K/m = 19.9 rad/s

lim H(j) if there are poles at the origin

H(j) is plotted on a linear-log scale. For

2. The gain in dB is therefore

2. The gain in dB is therefore

and described by parameters

, at all frequencies. The magnitude Bode

, at all frequencies. The magnitude plot is shown in below.

which is a horizontal line on the plot at 0dB (unity) gain.

The high frequency asymptote is found by assuming that 1,

and the high frequency asymptote is

1.1 Bode Plots of Higher Order Systems

2 radians/sec. and a damping ratio of

Das könnte Ihnen auch gefallen