0% found this document useful (0 votes)
17 views188 pages

Lecture Chapter1 8new

1. Modeling systems is important for analysis, design, and control. It involves making assumptions about elements, establishing relationships between elements, and obtaining mathematical descriptions. 2. Conventional modeling methods include schematic diagrams, which represent physical systems graphically, and constitutive relations, which define linear relationships between effort, flow, momentum and displacement variables in mechanical and electrical systems. 3. Schematic diagrams of examples like a cam and follower, vehicle suspension, and sewing machine can be used to derive governing equations. Constitutive relations define mass, damping, and spring parameters in mechanical systems and analogous components in electrical systems.

Uploaded by

Dinuka Ravimal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
17 views188 pages

Lecture Chapter1 8new

1. Modeling systems is important for analysis, design, and control. It involves making assumptions about elements, establishing relationships between elements, and obtaining mathematical descriptions. 2. Conventional modeling methods include schematic diagrams, which represent physical systems graphically, and constitutive relations, which define linear relationships between effort, flow, momentum and displacement variables in mechanical and electrical systems. 3. Schematic diagrams of examples like a cam and follower, vehicle suspension, and sewing machine can be used to derive governing equations. Constitutive relations define mass, damping, and spring parameters in mechanical systems and analogous components in electrical systems.

Uploaded by

Dinuka Ravimal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 188

Chapter 1

Introduction

1 Dynamic system modeling


Let me start with a question as shown in Fig1.1 why modeling is important. An object
is falling on the wooden table having legs, you are asked to describe velocity and dis-
placement motions of the table in the y direction after the object hits the table. What
is your first step to do for this question? What process is requried to take to undstand
the system behavior?

Figure 1.1: How to model a wooden table with legs

We need to think what elements the table is composed of. Is the upper part of the table
is thick of thin? Are the legs are regid or flexible? How are they treated? Is force applied

1
to the table or is velocity applied? What mathematical approach should be applied to
understand the motion? They must be other quesiotns to be answered in addition to
the above quesiotns.

For answering the several questions, we first need to make appropriate assumptions of
the elements on our own. When the upper part of the table is thick we can treat it
as mass. When the legs are flexible, we can assume that they are treated as springs.
Second, we establish modeling to show the interactiing relation of elements. The inter-
action can be related to each other through power or energy transformation. In other
words, the modeling process should provide information for all these questions, assum-
tions, and power transformation. Then, the modeling makes it easy to analyze and
design the system as we want. Lastly, a methmatical description is obtained from the
model in a quantative way. However, we should not much rely on this mathmatical
solution without understaning the physical concept. It is more important to understand
the physical concept in the solution rather than just obtain the matmatical solution itself.

In various engineering systems, there are governing equations which look differently from
various physical systems as in Eqns.(1.1), (1.2), and (1.3).

Newton eqn.(mechanical system)

F = mẍ + cẋ + kx (1.1)

Bernoullie eqn.(fluidic system)


ρ ρ
P1 + v12 + ρgz1 = P2 + v22 + ρgz2 (1.2)
2 2
Circuit eqn.(electrical system)
Z
di 1
V = L + Ri + idt, i = i1 + i2 (1.3)
dt C
These kinds of equations are usually used for system analysis. One thing to note here is
that there are typical different variables and parameters depending on physical systems.
Here, we have some questions regarding to the equations. Is it required to memorize
those equations for engineering applications without knowing where to apply? Is there
any similarities between them in terms of physical concepts. If any, how are they easily
observable and which structure can help understand them? In this book, we can find
out those answers using an graphical unified modeing approach, a bond graph. We shall

2
emphasize the similarities between the physical systems by means of a energy and power.
Thus your knowledge and intuition in one field can be used to strengthen your under-
standing other analogous fields. In Chapter 2, let’s first start with simple conventional
modeling appraoches.

Q: What is a model? – simplification of reality, but with understandable


approximation
Q: Why is a modeling process important? – Analysis, design, control

3
Chapter 2

Conventional modeling methods

1 Schematic diagram
One of graphical modeling methods is a schematic diagram. The advantage of this
method is that physical systems are represeneted graphically with their operation and
fucntions conceptually visualized. Examples of mechanical, electrical, fluidic system
as shown in Fig.2.1. The governing equations are obtained by applying the well known
principles of physics such as Newton’s law, electronic circuit thoery, Bernoullie equation,
etc..

Figure 2.1: Several physical systems represented using schematic diagrams

4
Figure 2.2: Schematic diagram of a cam and follower

In a mechanical system, Figure 2.2(a) shows a cam and follower that is used for gas
combustion valve in an automobile engine. Figure 2.2(b) shows its schematic diagram.
The governing equation is derived using a Newton’s law as

mf x¨f + cx˙f + (k1 + k2 )xf = k2 xc (2.1)

We can use Eqn.(2.1) for the analysis of the cam and follower. The design purpose of
this device is how to make xf follow xc as much as possible, namely xf ≈ xc .

5
Figure 2.3: Schematic diagram of a vehicle suspension

Figure 2.3(a) shows a 1/4 scale vehicle suspension that is used for absorbing an im-
pact from a rough road ground on automobile body. Figure 2.3(b) shows its schematic
diagram. The governing equation is derived using a Newton’s law as

mt x¨1 + c(x˙1 − x˙2 ) + k1 (x1 − x2 ) + k2 x1 = k2 x (2.2)


mv x¨2 + c(x˙2 − x˙1 ) + k1 (x2 − x1 ) = 0 (2.3)
We can use Eqn.(2.2) and (2.3) for the analysis of the vehicle suspension. The design
purpose of this device is how to suppress x2 for x as much as possible, namely x2 ≈ 0.

6
Figure 2.4: Schematic diagram of an earthquake seismograph

Q: Figure 2.4 shows an earthquake seismograph that is used for recording


a vibration signal due to earthquake. Draw its schematic diagram and de-
rive the governing equation.Think about how the output motion should be
against the input vibration

Figure 2.5: Schematic diagram of a door closer

Q: Figure 2.5 shows a door closer that provides a restoring force to close a
door when it is open. Draw its schematic diagram and and derive its gov-

7
erning equation.

Figure 2.6: Schematic diagram of a sewing machine

Q: Figure 2.6 shows a sewing machine where the blade cuts a wood piece.
Make appropriate assumptions needed and obtain its schematic diagram with
its governing equation.

2 Constitutive relations
2.1 Varaibles and parameters in a mechanical system
There are four variables in a mechanical system. force f , velocity v, momentum p, and
displacement x. f and v are called power variables because product f and v is a power.
p and x are called energy variables. We have relations of p and x from definitions as
df dx
p= , v= (2.4)
dt dt
Hence, only two variables among four are independent each other. For a reference, there
are also similar variables equivalent to f , v, p, and x in other physical systems such as
eletrical, fludic, magnetic systems, and so on as listed in Table 7.1. The definitions made
in a mechanical system can be also applicable to other systems. We can recognize there
are similarities among variables in those physical systems.

There are only 3 components in mechanical systems, those are mass m, damper charac-
teriszed by a damping coefficent c, and spring by a spring coefficent k. The definitions
of mass, damper, and spring can be respectively made using the four variables f , v, p,
and x. When damping force and spring force are indicated as fd and fs respectively, the
four variables are linearly represented as
p = mv, (2.5)

8
Table 2.1: Variables in physical systems

ef f ort(e) f low(f ) momentum(p) displacement(q)


Mechanical system f v p x
Electrical system e i λ q
Hydraulic system P Q Γ V
Magnetic system M φ̇ φ

fs = kx, (2.6)
fd = cv, (2.7)
In other words, the spring coefficent k is a linear paremeter that relates force to dis-
placement. The damping coefficent c is a linear paremeter that relates force to velocity.
Mass m is a linear paremeter that relates monentum to velocity. Equations(7.3), (7.4),
and (7.4) are called linear constitutive relations. However, those parameters can be
nonlinear as well as variables at a specific system such as a fludic system.

Q: Please check if the constitutive relations defined in a mechanical system


can be also found in electrical systems. Derive the corresponding equations
similar to Eqn.(7.3), (7.4), and (7.4).

2.2 Variables and parameters in an electrical system


There are four variables in an electrical system. voltage e, current i, momentum , and
displacement q. e and f are called power variables because product e and f is a power.
and q are called energy variables. We have relations of p and x from definitions as
de dq
λ= , i= (2.8)
dt dt
The definitions made in an electrical system can be also applicable to other systems.
We can recognize there are similarities among variables in those physical systems.

There are only 3 parameters in an electrical system, those are inductor L, resistor R, and
capacitor C. The definitions of an inductor, resistor, and capacitor can be respectively
made using the four variables e, i, λ, and q. When the voltage drop across a resistor

9
and capacitor are indicated as VR and VC respectively, the four variables are linearly
represented as

λ = Li, (2.9)
VR = Ri, (2.10)
1
VC = q (2.11)
C
In other words, an inductor L is a linear paremeter that relates monentum λ to current
i. A resistor R is a linear paremeter that relates voltage V to current i. The inverse of
a capacitor C1 is a linear paremeter that relates voltage V to charge q. Equations(2.9),
(2.10), and (2.11) are called linear constitutive relations.

3 Input sources: effort and flow sources


3.1 Input sources in mechanical systems
As shown in Fig. 2.2, and 2.3, the motion of dynamic systems is occured by input
sources. It is important to select a right input becuase wrong input selection make it
diffcult to analyze a system. Moreover, it will cause a unpredictable motion. There are
two types of sources as an input to a plant or system. They are generally called as an
effort and a flow sources. In a mechanical system, they are force and velocity.

Figure 2.7 shows again a schematic diagram of a cam and follower. We take the displace-
ment(or velocity input) xc as an input of the cam and follower because the displacement
can be applied without being affected by the reaction force due to the spring k2 . In other
words, the force applied to the spring k2 is strong enough to generate the displacemnt
xc . independently. Its governing eqquation is given by

mf ẍf + cẋf + (k1 + k2 )xf = k2 xc (2.12)

We might think that the velocity input in the device can be equivalently modeled using
a force input F under the reason that Eq.(2.12) can be transformed to

mf ẍf + cẋf + k1 xf = k2 (xc − xf ) (2.13)

The corresponding schmatic diagram can be drawn in Fig.2.8. Equation(2.13) looks like
an equivalent equation with a force input because k2 (xc − xf ) is apprently the same

10
Figure 2.7: Schematic diagram of a cam and follower modeled with a velocity input

as force F . However, F is not obtained without knowing the information xf that will
be known as a reult after F is applied. Hence, there is conflict that the output xf
should be known independently to determine the input F . Therefore, it can be said that
displacement input is a right choice in this case. When velocity is applied as an input,
It is applied directly to a spring not a mass. Similary, when force is applied as an input,
It is applied directly to a a mass not a spring.

Figure 2.8: Schematic diagram of the cam-follower device equivalently modeled with a
force input

How about the input applied to a vehicle suspension shown in Fig.2.3? Which type of
an input is more appropriate? The answer is displacment(or velocity) because it is more
easily available from a road surface in a real application. Hence, the displacement input
x is applied directly to the spring. Another way to determine an input source is whether
motion takes place in a contact or noncontact with an input soruce. When mechancal
motion takes place in a contact with an input, the source is displacemnt(velocity). When

11
mechancal motion takes place in a noncontact with an input, the source is force. One
example of a force input can be found in an electrical motor becasue force is generated
by the electromagnetic principle in a noncntact way.

Q: Which input source modeling is more appropriate for the inch worm motor
driven by a PZT actuator? Suppose the gripping force 10N , the friction
coefficient 0.5, the gripping and detaching frequency 100Hz. Then, how fast
the shaft will be moving forward? One step moving distance is 0.1mm. The
mass of the shaft is 100gr.

Figure 2.9: Inch worm motor

Q: Figure 2.10(a) and (b) show the configuration of a DC motor and a step
motor. What is the difference between a DC motor and a step motor in
terms of input source type?

12
(a) (b)

Figure 2.10: Configuration of (a) a DC motor and (b) a step motor

3.2 Input sources in electrical systems


In electrical systems, there are also the same effort and flow sources as mechanical
systems, that is, voltage and current. What are the conditions for an ideal voltage and
current soruces in an electrical system?

3.2.1 Ideal voltage source

Figure 2.11: Ideal voltage source

An ideal voltage source is defined as a voltage supply that applies the input voltage to
a load without having a voltage loss as shown in Fig8.11. Therefore, the internal serial
resistance Rin should small enough to be ignored, that is, Vi n = 0 so that V is fully
applied to RL . Here, note that the load voltage VL is measured using a oscilloscope that
has an ∞ impedance in order not affect the load RL . If the impedance of the oscilloscope
is similar to RL , VL would be smaller than V , which is not intented to be.

13
3.2.2 Ideal current source

Figure 2.12: Ideal current source

An ideal current source is defined as a current supply that applies the input current to a
load without having a current loss as shown in Fig8.13. Therefore, the internal parallel
resistance Rin should be ∞ enough to be treated an open circuit, that is, Iin = 0 so that
I is fully applied to RL . Here, the load current IL is measured using a current meter
that has an 0 impedance.

3.2.3 Practical voltage source

Figure 2.13: Practical voltage source

If RL = Rin , VL is much reduced compared to the case RL  Rin . In other words, the
voltage source V is not fully applied to the load voltage VL , which indicates it is not an
ideal voltage source. This kind of voltage drop in an electrical system is rarely observed
because RL is usually much larger than Rin . However, this effort drop in the source is
easily observed in mechanical or hydraulic systems. This problem leads to a situation
of lack of flow(velocity) in the load.

14
4 Load Curve
Though it was addressed in 3.1 that the force source and flow source are applied to a
mass and spring respectively, these sources are not actually practical in real systems.
Figure 2.14 shows a typical Toque T versus angular velocity ω curve of an automobile
engine in which the product of T and ω is constant. Here Tst is the stall torque, a
maximum torque obtained at zero agnular velocity. ωnl is no-load angular velocity, a
maximum angular velocity obtained at zero torque.

Figure 2.14: T − ω curve of an automobile engine

Tst is decreased when motion takes a place. The decreased torque can be determined by
the load curve characterized by T (ω). On the other hand, the decreased angular velocity
can be determined by the load curve characterized by ω(T ). Hence, the practical effort
source is represented using Tr as showin in Fig 2.15(a). The practical flow source is
also represented using ωr as showin in Fig 2.15(b). When the load is linearly varying,
the relations of the load can be T (ω) = Rω or ω(T ) = T /R respectively, where R is
resistance of the load.

15
Figure 2.15: Input sources represetned by either (a) an effort soruce or (b) a flow source
in a practical system.

So, the practical effort and flow inputs are mathematically represnted as

Tr = Tst − T (ω) = Tst − Rω (2.14)


ωr = ωnl − ω( T ) = ωnl − T /R (2.15)

16
5 Time response analysis for the 2nd order system
When a governing equation is obtained, the next step is to solve the differential equation
to undedstand the system behavoir for a specific input. We study on most popular
mechanical system composed of mass, damper, and spring with a force input as shown
in Fig.(2.16). Its governing equation is represented in a second order differential form as

Figure 2.16: 2nd order mechanical system with a force input

F = mẍ + cẋ + kx (2.16)

Here, m, c, and k are the mass, damping coefficient and spring coefficent of the me-
chanical system. Eq.(2.16) is transformed to Eq.(2.17) using the Laplace transformation
method under the assumption of ẋ = ẍ = 0.

F (s) = (ms2 + cs + k)X(s) (2.17)


Here, we introduce the transfer function of a linear, time invariant, differential equation
system defined as the ratio of Laplace transformation of the output(responce function)
to the Laplace transformation of the input(driving function) under the assumption that
all initial conditions are zero. Then, the transfer function of output displacement x(t)
versus input force F (t) is obtained as

X(s) 1
= 2
(2.18)
F (s) ms + cs + k

Equation(2.18) is rewritten for normaization as

kX(s) k k/m
= = 2 (2.19)
F (s) 2
ms + cs + k (s + mc s + k
m
)

17
k
Define ωn2 = m and 2ζωn = mc , then ζ = 2√cmk . ωn is called a natural frequency. ζ is
called a damping ratio. Using more genral notations of output C(s) to input R(s),

C (s) ωn2
= 2 (2.20)
R (s) s + 2ζωn s + ωn2
The dynamic behavior of Eq.(7.52) is described in terms of two parameter ζ and ωn . De-
pending on the range of ζ, the output resonse is called underdamped, critically damped,
overdamped.

Underdamped case ( 0 < ζ < 1 )

The roots of the characteristic eqn. s2 + 2ζωn s + ωn2 = 0 are


p
s = −ζωn ± jωn 1 − ζ 2 = −ζωn ± jωd = σ ± jωd (2.21)

Frm the above definitions of σ and ωd , the relation of ωn and ζ are graphically obtained
in Fig. 2.17. ωn 2 = σ 2 + (jωd )2 , ζ = cosβ

Figure 2.17: Defintiond of ωn and ζ

Set s1 = −ζωn − jωd and s2 = −ζωn + jωd , then

C (s) ωn2 ωn2


= 2 =
R (s) s + 2ζωn s + ωn2 (s − s1 ) (s − s2 )
2
ωn
=
(s + ζωn + jωd ) (s + ζωn − jωd )

18
R∞ h i∞
−st e−st
For a unit step input R(t) = 1, R(s) = 1/s (ref. L {1} = 0
(1e ) dt = −s
= 1s ).
0

ωn2 1 s + 2ζωn
C (s) = 2 2
= − 2
(s + 2ζωn s + ωn ) s s (s + 2ζωn s + ωn2 )
1 s + ζωn ζωn
= − −
s (s + ζωn ) + ωd (s + ζωn )2 + ωd2
2 2

It is shown that
( )
−1 s + ζωn
L = e−ζωn t cos (ωd t)
(s + ζωn )2 + ωd2
( )
−1 ωd
L = e−ζωn t sin (ωd t)
(s + ζωn )2 + ωd2

Hence,
ζ
L−1 C(s) = c(t) = 1 − e−ζωn t (cos(ωd t) + p sin ωd t)(2.22)
1 − ζ2
p !
e−ζωn t 1 − ζ2
=1− √ sin ωd t + tan−1 (2.23)
1−ζ ζ
As written in Eq.(2.23), the output c(t) has a steady state response and transcient
response. It can be seen that the frequency of transcient oscillation is the damped
natural frequency wd and thus varies with the damping ratio. It is drawn in Fig. 2.18.

The transcient response of a practical system often exhibit damped oscillations before
reaching steady state. In specifying the transcient response charachteristics of the 2nd
order system to a unit step input, it is common to specify the following.
1. Delay time, td
2. Rise time, ts
3. Peak time, tp
4. Maximum overshoot, Mp
5. Settling time, ts

These specifications are grphically defined in Fig. 2.19. Referring to Eq.(2.23), we ob-
tain the rise time tr by letting c(tr ) = 1.
The settling time is the time required for the response curve to reach and stay within

19
Figure 2.18: Output response of the 2nd order mechanical system

a range of about the final value of size specified by absolute percentage of the final
value(usually 2 or 5per)

As addressed in the above, the time domain analysis provides an accurate result. How-
ever it is very complicated and tedious process because the differential equation needs to
be solved anyway. Moreover, an output responses should be calculated again when an
input and order of plant model change. Therefore, it can be said that the time domain
analysis is not a good tool to understand a system behavior in terms of an engineering
sense. The study of follwoing cases of critically damped and overdamped are left to
readers who are intereseted in completeness of mathematical analysis for your refernces.

Critically damped case ( ζ = 1 )

If the two poles of C(s)/R(s) are nearly equal, the system may be approximated by a
critically damped one. For a unit-step input, R(s) = 1/s and C(s) can be written

20
Figure 2.19: Unit step response curve showing specifications

ωn 2
C(s) = (2.24)
(s + ωn )2 s
The inverse Laplace transform of Equation (2.24) may be found as

c(t) = 1 − e−ωn t (1 + ωn t) (t ≥ 0) (2.25)

This result can be obtained by letting ζ approach unity in Equation () and by using the
following limit: p
sin ωd t sin ωn 1 − ζ 2 t
lim p = lim p = ωn t
ζ→∞ 1 − ζ 2 ζ→∞ 1 − ζ2

Overdamped case ( ζ > 1 )

In this case, the two poles of C(s)/R(s) are negative real and unequal. For a unit-step
input, R(s) = 1/s and C(s) can be written

ωn 2
C(s) = p p (2.26)
(s + ζωn + ωn ζ 2 − 1)(s + ζωn − ωn ζ 2 − 1)s

21
The inverse Laplace transform of Equation (2.26) is
1 √
2
c(t) = 1 + p p e−(ζ+ ζ −1)ωn t
2 ζ 2 − 1(ζ + ζ 2 − 1)
1 √
2
− p p e−(ζ− ζ −1)ωn t
2
2 ζ − 1(ζ + ζ − 1)2

ωn e−s1 t e−s2 t
=1+ p ( − ) (t ≥ 0) (2.27)
2 ζ 2 − 1 s1 s2
p p
where s1 = (ζ + ζ 2 − 1)ωn and s2 = (ζ − ζ 2 − 1)ωn . Thus, the response c(t) includes
two decaying exponential terms.

22
6 Frequency response analysis
Consider a stable linear time-invariant system whose transfer function is G(s) as shown
in Fig.7.12. The input and output are denoted by x(t) and y(t) respectively. If x(t)
is sinusoidal signal, the steady state output wiil be also sinusoidal signal of the same
frequency but with possibly different magnitude and phase angle

Figure 2.20: Stable linear time-invariant system

Let’s assume that the input signal is given by

x(t) = Xsinωt (2.28)

Then, the output singal can be represented by

y(t) = Y sin(ωt + φ) (2.29)

Suppose that the transfer function G(s) can be written as a ratio of two polynominals
in s as
p(s) p(s)
G(s) = = (2.30)
q(s) (s + s1 )(s + s2 ) · · · (s + sn )
The Laplace transformed ouput Y (s) is then

p(s)
Y (s) = G(s)X(s) = X(s) (2.31)
q(s)

where X(s) is the Laplace transform of x(t).

It was proven that the frequency response of characteristics of a system can be obtained
directly from the sinusoidal transfer function, that is, the transfer function G(jw) in

23
which s is replaced by jw, where w is freqeuncy[?]. It can be recalled that σ is only
related to tranciernt response. Hence, the steady state response can be given by
Y (jω)
G(jω) = | | = M elφ = M 6 φ(2.32)
X(jω)
where M is the amplitude ratio of the output and input sinusoids and φ is the phase
shift between the input sinusoid and output sinosoid. In the frequency response test,
the input frqeuency w is varied until entire frequency range of interest is covered. An
example of input and output sinusoidal signals is shown in Fig.7.13

Figure 2.21: Input and ouput sinusodial signals having different amplitude and phase
delay

Now, let us find out the transfer function of the 2nd order mechancal system expressed
Eq.(7.52). The transfer function G(jω) is represented as

C(iω) ωn2
G(jω) = = 2 (2.33)
R(jω) ωn − ω 2 + j(2ζ ωωn )
The amplitude ratio M and phase delay φ of G(jω) are respectively obtained as
1 1
M = |G(jω)| = = (2.34)
1− ( ωωn )2 + j2ζ ωωn a + bj
1
φ=6 ( ) (2.35)
a + bj
M and φ are are described in Fig.7.19. They are described in a logarithm scale Bode
plot with respect to applied frequency.

24
Figure 2.22: Frequency response of the 2nd order system

6.1 Specifications of steady state response characteristics


labelSpecifications of steady state response characteristics If G(jw) has a peak value
at some freqeuncy, this frequency is called the resonant frequency wr . A peak value of
|G(jw)| will occur when
w2 w
g(w) = (1 − 2 )2 + (2ζ )2 (2.36)
wn wn
is a minimum. Thus, the resonant frequency wr is
p
ωr = ωn 1 − 2ζ 2 (2.37)

25
The magnitude of the resonant peak Mr can be found by substituting Eq. (7.58) into
Eq. (7.54). For 0 ≤ ζ ≤ 0.707,

Mr = |G(jw)|max = |G(jwr )| = √1 (2.38)


2ζ 1−ζ 2

φ
tp = (2.39)
wd

4
ts = , (2%criterion) (2.40)
wn
3
ts = , (5%criterion) (2.41)
wn

6.2 System bandwidth


The system bandwidth is defined a maximum frequency at which the magnitude of
the transfer function is 3db below its zero-frequency value. It is oftten called a cutoff
frequency.This kind of filter is

— Control point of view

— Signal point of view

The bandwidth of the plant is regarded as a speed performance index in a mechanical


system.
Q. Draw the time response of the plant when ω = ωl (low frequency), ω = ωn ,
and ω = ωh (high frequency).

We can find real examples of plant whose behavior can be understood using the fre-
quency response analysis
– car driving
– airplane landing

Q. How are the magnitude ratio and phase delay of a system shown in Fig.
2.2?

26
6.3 The advantages of Laplace transformation
The topic of Laplace transformation is covered in several books [?, ] Here, some impor-
tant things in the engineering field will be addressed for summary.

R∞ h i∞
−st e−st 1
— L{1} = 0
(1e )dt = −s
= s
0
R∞
— L {At} = 0 (Ate−st ) dt = sA2
n o R  
∞ df (t) −st
— L df(t)
dt
= 0 dt
e dt = F (s) · s
nR o R∞
t Rt F (s)
— L 0
f (τ ) dτ = 0
(e−st ) 0
f (τ ) dτ dt = s

— Easy understanding of system characteristics

— Exact(explicit) representation of a system using a transfer function For example,


it is impossible to express a transfer fucntion of output displacement x for force
input F , Fx(t)
(t)
for a system described in

F = mẍ + cẋ + kx. (2.42)


X(s)
However, it is possible to express the transfer fucntion F (s)
easily because F (s) is
represented n a form of G(s)X(s)
A(s)
— Multiplication of A(s) · B(s) and division B(s) turns to addition A(s) + B(s) and
subtraction A(s) − B(s) in a Bode plot or in a logarithmic plot

6.4 Frequency responses of some typical plants


Let us investigate how the transfer fucntions of mechanical sytems are reprenseted in
the following cases.

— Mass only

— Mass and damper

— Mass, spring, and damper with a force input and velocity output

— Mass, spring, and damper with a velocity input and velocity output

— 1/s, s, 1/(s+100), s+100, 1/s(s+100)

27
When a force is applied to a unit mass, the velocity is obtained by integrating the force.
It can be represented using a block diagram as showin in Fig.2.23. Since the Laplace
Rt
transformation of time integration 0 f (t)dt is expressed as F (s)/s, the transfer function
of time integration is 1/s as shown in Fig.2.23. The 1/s plant is called an integrator.
Hence, a constant force input causes a drift and eventually makes a mechanical plant
collide at the end. When a sine fucntion is applied to the integrator a cosine signal
is produced in an ideal case. However, a real ouput signal has a drift because of a
low frequency disturbance included in the applied force and the DC offset produced in
electronic components

Figure 2.23: Laplace transformation of time integration

Q: How does a mechaical system composed of a mass and damper behave


when a sinusoidal force is applied to a unit mass.

The transfer fucntions of the mechanical systems are also easily implemented using
electronic circuits such as an integrator, differentiator, and low pass filters. It can be
said that the 2nd order mechanical plant is nothing but the 2nd order low pass filter. For
example, let us find out a transfer fuction Vo /V1 of the electronic circuit shown in Fig.8.8
made of an operational amplifier. It is known that the differential equation of the circuit
is derived using the characteristics of the OP amp, that is, a high input impedance and
a low output impedance as Z
1
v0 = v1 dt (2.43)
RC
Hence, the transfer fucntion is
Vo (s) 1
= (2.44)
V1 (s) RCs
It is an integrator whose frequency response is shown in Fig.2.25 in a Bode plot

28
Figure 2.24: An electronic circuit made of an operational amplifier

Figure 2.25: A Bode plot of an integrator

Q: Explain why 1s , or s12 plant have the magnitude reduction effect at a higher
frequency?

Q: Explain why a contant force is not integrated in a low frequency for a


1
plant whose transfer function is s+100

Q: Draw the Bode plot for the following transfer function:

10(jw + 3)
(2.45)
jw(jw2 + jw + 2)

Q: How is the output position signal when a low frequency command signal(or
DC offset) and high frequency noise are applied to a plant of a mass and
damper system?

29
6.5 Various time inputs to a plant
It is also important to understand a time response from the frequency response using
the magnitude ratio M and phase delay φ. It is possible because the output response
can be expected because M and φ can be calculated for varying frequency ω.

Figure 2.26: Responses from some inputs

The validation of frequency response analysis for various inputs such as impulse,
step, ramp inputs are made by the fact that those inputs can represented by series of
sinusoidal functions, that is Fourier series representation.

6.6 Good Plant Design


What is a good design of a plant in terms of good tracking performance? It is a plant
with no resonance in a high frequency as shown in figure 2.27
It is because of difficulty in control of the plant at a high frequency.

Figure 2.27: Considering points to make a good system

7 Passive system; non-isolated system


There is a mechanical system composed of mass(m1 ), damper(c1 ), spring(k1 ) with force
input as shown in Fig.2.28(a). Its governing equation is obtained and described in

30
2nd order differential equation. If another mechanical system composed of mass(m2 ),
damper(c2 ), spring(k2 ) is serially connected to the previous system to have a 2 degree-
of-freedom mortion as shown in Fig.2.28(b), can each system be described seperately
like two isolated systems?

Figure 2.28: Schmatic diagram of mechanical system with (a) 1 degree-of-freedom (b) 2
degree-of-freedom

We repeat this question for an electronic circuit shown in Fig.8.1 for analysis simplicity
. The dynamics of the electronic circuit composed of R1 and C are represented as

Figure 2.29: An electronic circuit composed R1 , C, and R2

31
Z
1 L{v(t)} 1
1) v (t) = R1 i (t) + i1 (t)dt −−−−→ V (s) = R1 I (s) + I1 (s)
C Cs
L{i(t)}
2) i (t) = i1 (t) + i2 (t) −−−−→ I (s) = I1 (s) + I2 (s)
Z
1 L{vo (t)} 1
3) vo (t) = i1 (t)dt = R2 i2 (t) −−−−−→ Vo (s) = I1 (s) = R2 I2 (s)
C Cs

1
Vo I (s)
Cs 1 Vo 1
= 1 =⇒ =r
V R1 I (s) + Cs I1 (s) V h i2
[R1 Cω]2 + 1 + R1
R2
1
I
Cs 1
(s)
= 1
R1 [I1 (s) + I2 (s)] + I
Cs 1
(s)
1
= h i
R1
R1 Cs + 1 + R2

Figure 2.30: Transfer function of the electronic circuit

As shown in Fig.2.30, the transfer function Vo /V , is not only dependent on R1 , C but


also dependent of R2 , which indicates that Vo /V is not an isolated system. In conclusion,
it is unavoidable to have a non-isolated system if a system is passive circuit. In order to
have an isolated system, a system should be feedback system(servo system) such as an
op-amp.

32
Chapter 3

Bond graph modeling

1 Bond graph modeling in mechanical systems


It was stated in Ch. 2 that there were only 3 components in mechanical systems, those
are mass m, damper represented by a damping coefficent c, and spring represented by
a spring coefficent k. There are no other elements except these three elements. There
are four variables, force F , velocity v, momentum p, and displacement x. F and v are
called power variables because product F and v is a power. p and x are called energy
variables. We have relations of p and x from definitions as
dF dx
p= ,v = (3.1)
dt dt
Hence, only two variables among four are independent each other.

The definitions of mass, damper, and spring can be respectively made using the four
variables F , v, p, and x.They are linearly represented as

p = mv (3.2)

F = kx (3.3)
F = cv (3.4)
In other words, the spring coefficent k is a linear paremeter that relates force to displace-
ment. The damping coefficent c is a linear paremeter that relates force to velocity. Mass
m is a linear paremeter that relates monentum to velocity. Equations(3.2), (3.3), and
(3.4) are called linear constitutive relations. However, those parameters can be nonlinear

33
at specific systems. Furthermore, the constitutive relations can be also nonlinear.

Q: When force is expressed as in Eqn.(3.5), what kind of costitutive relation


is the force?

F = c2 v 2 (3.5)

The mechanical energy E is defined by integrating the mechanical P . E is represented


using the constitutive relations as
Z Z Z Z
dx 1
E = P dt = F vdt = F dt = F dx = kx2 (3.6)
dt 2
p2
Z Z Z Z
dp 1
E = P dt = F vdt = v dt = vdp = = mv 2 (3.7)
dt 2m 2
Since the mechanical power P is a product of force F and velocity v, F and v are called
power varaibles. Similarly, since the mometum p and displacement x are appeared in the
energy relations as indicated in Eqn.(3.7) and (3.7), p and x are called energy variables.

34
1.1 Passive 1-Ports: m, c, k
When the input is applied to a mechanical system composed of mass, spring, and damper,
the input energy is delivered and dissipated in the elements. We call them passive 1-port
because they exchange power(or energy) at a single location or port and they contain
no sources of power. The bonds with half arrows are used to indicate the direction of
the power or energy.

Figure 3.1: The passive 1-ports of a mechanical system

1.2 Junctions
1-junction(common velocity junction)

Figure 3.2(a) shows the schematic diagram of a mechanical system that looks like a cart
on rollers where three forces are acting. The cart is assumed massless. Figure 3.2(b)
shows a 3 port 1 junction diagram to express the velocity of the three points of force
acting on the cart is the same. The relations associated with the 1 junction are
F3 − F2 − F1 = 0
v1 = v2 = v3 = V

The half arrows point toward the 1-junction for power supplied and away for power
delivered. Example system using a 1-junction is as shown in Fig.

35
Figure 3.2: (a) A cart on rollers where three forces are acting, (b) 3 prot 1-junction
diagram

Figure 3.3: Example : 1-juntion

0-junction(common force junction)

Figure 3.4 (a) shows a damper connected between two carts. Figure 3.4(b) shows a 0
junction diagram to express the force acting on the damper is the same. The relations
associated with the 1 junction are

v1 = v2 + v3
F1 = F2 = F1 = F

Figure 3.5 show a typical example of a mechanical system composed of a mass, spring,
and damper. Figure 3.6 shows its bond graph. A bond graph shows how the power or
energy associated with the parameters m, c, and k are represented and delivered in a
system. The bonds with half arrows are used to indicate the direction of the power or
energy. First, the force F is applied to the mass m and it is represneted as an effort
source in the bond. Second, the net force F − F1 acting on the mass m is represetned
in the bond using I-port. Then, the velocity v of the mass is obtained as a result of the
net force applying to the mass m. It can be sadi that v is obtaned from the mometum
R
p divided by m because of the relation of p = (F − F1 )dt. Here, we can know the
momentum p is just integral of force difference on the mass. Third, the force F1 with
velocity v is delivered to other components R and C ports. Lastly, when the velocity

36
Figure 3.4: A damper connected between two carts

v is applied to the damper and spring, the damping force Fd and spring force Fs are
respectively determined using the constitutive relations.

Figure 3.5: Bond graph of a mechanical system composed of mass, spring, and damper

Here, 1-juction is used to indicate that the bonds have the same velocity. Force sumation
of in-going bonds are the same as force summation of outgoing bonds. Velocity sumation
of in-going bonds are the same as velocity summation of outgoing bonds. The 1 junctions
connected in a row can be equivalently expressed as shown in Fig. 3.7

Figure 3.8(a) shows an example of a mechanical system composed of a mass, spring,


and damper connected in series. Figure 3.8(b) shows its bond graph model showing how

37
Figure 3.6: Bond graph of a mechanical system shown in Fig. 3.5

Figure 3.7: Equivalent bond graph of a mechanical system shown in Fig. 3.6

the force and velocity are delivered through the components using the 1 juctions and 0
junctions

This bond graph modeling method takes many advantages compared with other mod-
eling methods. First, we can recognize easily which components are used for modeling
because they are graphically depicted at ports. Secondly, it shows which input type is
chosen between force and velocity. Thirdly, it shows how force or power are distributed
and consumed at the I-port, R-port, and C-ports. In other words, force or power will
be eventually decreased as they are encountered with several ports because the mass,
damper, and spring are acting as loads against the applied force or power.

38
Figure 3.8: (a) A mechanical system composed of a mass, spring, and damper connected
in series and (b) its bond graph

Q: Draw bond graphs of the mechanical systems shown in Fig. 3.9, 3.10,
3.11, 3.12, 3.13

39
Figure 3.9: Example 1

40
Figure 3.10: Example 2

Figure 3.11: Examples 3

41
Figure 3.12: Example 4

Figure 3.13: Example 5

42
2 Causality
Causality is determination of input and output variables of elements on a bond graph.
The primary restriction of causality can be stated as

“Either the effort or flow must be for a port but both cannot be input or output si-
multaneously on the same port.” It is impossible to simultaneously specify the force
and velocity as shown in Fig 3.14. For example, when 1N is applied to a mass 1kg
the velocity is determined in return by integrating the force with respect to time. In
other words, It is not possible to apply force 1N to a mass 1kg and obtain the velocity
10m/sec at the same time.

Figure 3.14: Causallity of sources

If the effort is an input to a component B then the flow must be an output and vice
versa. Effort is indicated using a bar attached to an arrow. The bar and point on the
half arrow are graphically understood by visualizing ”Effort pushes and flow points”.

Figure 3.15: Indication of effor and flow by using a bar and a point

43
2.1 Source causality
By element definition, the proper causality of sources are shown

2.2 Causalities of energy storaging elements


For energy source elements, namely C-port and I-port elements, there are two types of
causality. Consider the C element The cosntitutive law for a C always relates e(ef f ort)
to q(displacement), and q is the integral of f (f low). Thus, if f is the input to a C, it
is first integrated to find q and then e is obtained in return as a function of q(phis (q)).
That is why this C port is called an integral causality. This situation in graph and
equation form is shown in Fig. 3.16

Figure 3.16: Integral causality C-port

If e is the input to a C, q is first obtained by a fucntion of φ−1


s , then f is obtained in
return by differentiating q. That is why this C port is called an derivative causality.
This situation in graph and equation form is shown in Fig. 3.17

Figure 3.17: Derivative causality C-port

The same concepts apply to the I element. The I element relates f to p, and p is
the time integral of e. If e is the input to a I, p is obtained by integrating e, then f
is obtained in return as a function of φs . That is why this I port is called an integral
causality I port. This situation in graph and equation form is

44
Figure 3.18: Integral causality I-port

If f is the input to a I, p is obtained as a function of φ−1


s , then e is obtained in return
by differentiating p. That is why this I port is called a derivative causality I port. This
situation in graph and equation form is

Figure 3.19: Derivative causality I-port

When there are derivative causalities in mechanical structures, they would experi-
ence a large impact, big oscillation, saturation which eventually causes its breakdown.
For example, if a step flow input is applied to at an I element, infinity force can be
produced in return. Hence, we should take caution in designing mechanical systems so
that they have no derivative causalities in I and C elements.

2.3 Causality of energy dissipating element


Causality for R can be imposed in either direction as shown in Fig. 3.20. When an
effort e applies to R, the flow f is determined in return and when a flow f applies to R,
the effort e is determinded in return.

Figure 3.20: Causality for R port

45
2.4 Causality of junction elements
The 0− and 1− junctions have three possible causal patterns if they are 3-ports. We
consider a specific example

e1 = e2 = e3
f1 − f2 − f3 = 0

Suppose that f1 and f3 are impressed on the 0 juction as inputs, then, the third flow
must be an output becasue of the flow relation. Hence, the third effort must be deter-
mined in return. In conclusion, there should be only one bar at the bonds directing to
the 0 junction. The resulted force is transmitted.

Q: How is the causality bond when e1 is applied to 0 junction having 3 ports?

The 1 junction causal pattern are found by the same kind of reasoning, with efforts
and flows interchanged. As an example, at the 1 junction efforts e1 and e3 are applied,
the difference of effort (e1 − e3 ) is applied to I. Then, velocity is determined in return.
In conclusion, there should be only one point at the bonds directing to the 1 junction.
The resulted velocity is transmitted.

f1 = f2 = f3
e1 − e2 − e3 = 0

Q: Draw a bond graph of trains connected by rigid links shown in Fig. 3.21
when a step force is applied to the first train. What will happen to trains?
How are they modified to avoid mechanical problems.

46
Figure 3.21: Train connected by rigid links.

3 State equations
Bond graphs lend themselves to state space methods, in which an nth order dynamic
system is represented by n first-order differential equations in n variables. The state
equations are closely connected to the energy stored in the system. Variables in bond-
graph equations are sets will always be the energy variables p and q, and the state
equations are effort ṗ and flow q̇ equations. Hence, the number of the state equations is
the same as the number of the integral causalities in the bond graph.

47
Figure 3.22: State equation for the system having two mass and two springs

As you can see, the n first order equations are generally simpler in form than the
smaller number of higher order equations. Furthermore, from the definition of ṗ1 =
m1 ẍ1 , X1 = x − x1 , ṗ2 = m2 ẍ2 , and X2 = x1 − x2 , we can see the state equations are
the same equations that are obtained by applying Newton’s method.

48
Figure 3.23: Trains having a derivative causality

Let’s think about how the state equation is obtained when the derivative causality
exists in the bond graph. There is a train connected with a rigid link shown in Fig.
3.23. From its bond graph we know there is a derivative causality. Since there is one
integral causality, there must be one state equation. The equation are derived using the
constitutive relations as

P1
Ṗ1 = F − F1 : v = (3.8)
m
F1 + F = Ṗ2 : Ṗ2 = mv̇ (3.9)

Finally, the state equation is obtained as

Ṗ1 = F − Ṗ2 + F = 2F − mv̇ = 2F − Ṗ1 (3.10)

Hence, we have Ṗ1 = F .


We can obtain the state equation regardless of derivative causality existence. We
take an advantage of knowing whether a system has a causality problem or not from .
If it has a derivative causality, we know that it will cause unstable operation.

49
Q: Draw a bond graph of the mechanical structure shown in Fig. 3.24 and
derive its state equation.

Figure 3.24: A mechanical structure composed of mass, spring, and transformer

50
4 Bond graph modeling in electrical systems
It was alredady stated in Ch. 2 that there were only 3 components in electrical systems,
those were an inductor L, a resistance R, and a capacitor C. There are also four
variables, voltage e, current I, flux linkage λ, and charge q that are equivalent to force
f , velpcity v, momentum p, and displacement x of the mechanoical system. V and i are
called power variables because product e and i is a power. We have relations of λ and i
from definitions as.
de dq
λ = ,i = (3.11)
dt dt
Hence, only two variables among four are independent each other.

The definitions of an inductor, resistor, and capacitor can be respectively made using
the four variables e, e, λ, and q.They are called constitutive relation represented as

λ = Li (3.12)
1
e= q (3.13)
C
e = Ri (3.14)
In other words, the inductor L is a linear paremeter that relates the electrical monen-
tum to current. The inverse of the capacitor is a linear paremeter that relates voltage
to charge. The resisot is a linear paremeter that relates voltage to current.

The electrical energy E is defined by integrating the electrical power P with respect
to time. Hence, E is represented using the constitutive relations as
Z Z Z Z
dq 1 2
E = P dt = eidt = e dt = edq = q (3.15)
dt 2C
λ2
Z Z Z Z
dλ 1
E = P dt = eidt = i dt = idλ = = Li2 (3.16)
dt 2L 2
Since the electrical power P is a product of force e and velocity i, F and v are called
power varaibles. Similarly, since the mometum and displacement q are appeared in the
energy relations as indicated in Eqn.(7.12) and (7.12), λ and q are called energy variables.

51
Figure 7.8 shows an example of an electronic circuit composed of L, R, and C.

Figure 3.25: An example of an electronic circuit

Figure 7.9 shows its bond graph representation. shows how the power or energy
associated with the parameters L, R, and C are represented and delivered in the sys-
tem. The psssive 1-ports for the three components and 1 and 0 juntions are shown. We
call them passive 1-ports because they exchange power(or energy) at a single location or
port The bonds with half arrows are used to indicate the direction of the power or energy.

Figure 3.26: of the electronic circuit shown in Fig. 7.8

First, the voltage input e is the applied to the electric circuit and it is represneted as
an effort source in the bond. Second, the net voltage e − e2 acting on the resistor R1
is represetned in the bond using 1 port. Then, the current i1 is obtained as a result
of the net voltage applying to the resistor R1 . Third, the voltage e2 with current i1 is
delivered to the next component C port. Fourth, i1 is divided into two different paths
reuslted in providing i2 and i3 . i2 is applied to the C port and generates the voltage e2

52
using the constitutive relation qC2 . Lastly, e2 is applied to the next element R2 . The net
voltage e3 is acting on the R2 element is represnted. Then i3 is obtained as a result. The
resistor voltage e3 and e4 are respectively determined using the constitutive relations.
One thing to note is that the inductor, resistor, and capacitor are acting as loads to the
applied voltage or power. Hence, voltage or power will be eventually decreased as they
are encountered with these kinds of loads.

Here, 1-juction is used to indicate that the bonds have the same current. voltage
sumation of in-going bonds are the same as voltage summation of outgoing bonds. Cur-
rent sumation of in-going bonds are the same as current summation of outgoing bonds.
In the 0-junctions associated with the C port we have the relations of

i1 = i2 + i3 (3.17)
In the 1-junctions associated with the second R port we have the relations of

e2 = e3 + e4 (3.18)
Q: Draw a bond graph of the electrical circuit shown in Fig. ?? and derive
its state equation. Explain how the flux linkage is eaxily understood
.

simple_circuit.png

Figure 3.27: simplec ircuit

53
4.1 Causality problem in an electronic circuit
Figure 3.28 shows an electronic circuit which does not perform its function. This example
shows how the eletric circuit shoud be modified to make it operate functionally using
the causality concept. The OP amplifier is used to intend to supply a sinusoidal current
to the laser diode with a ± swing of a desired amplitude. The voltage divider are used
to add a constant positive current in order to flow a positive sinusoidal current at the
LD.

Figure 3.28: Causality problem in an electronic circuit

54
According to the fact the OP amplifier is a voltage source, an effort source is denoted
using a bar at the direction to the 0 junction. Next, the bond leaving from the 0 junction
should have a flow causality denoted using a point at the direction to the LD becauise
it is our design intention of the electric circuit. Then, there a causality violation at the
0 junction as seen in Fig. 3.28. To avoid this situatiuon, we need to modify it as shown
in Fig. 3.29. One of appropriate electric circuts corresponding to Fig. 3.29 is shown in
Fig.3.30.

Figure 3.29: ModifiedCausality

Figure 3.30: ElectricCircuitWithLoad

A resistor or a capacitor should be inserted after the OP amp as loads to generate


current from the OP amplifier. The causality mark in the first R element indicates that
the voltage difference is applied to generate current i1 in return. Without the R, it is
not possible to produce i1 .

55
4.2 State equations of electrical circuits
Similary in the mechanical system, the state equations are closely connected to the
energy stored in the system. Variables in bond-graph equations are sets will always be
the energy variables and q, and the state equations are ˙ and q̇ equations. The number
of the state equations is the same as the number of the integral causalities in the bond
graph.
For a bond graph shown in Fig. 3.31, the state equations are deriven in Eqn.

Figure 3.31: StateEquationOfElectricalCircuits

λ
λ̇ = E − eR1 − eC : eR1 = R1 i, i = (3.19)
L
eC 1
q̇ = i − i2 : i2 = , eC = q (3.20)
R2 C
Q: Derive the state equation for the electric circuit shown in Fig.3.32

Figure 3.32: StateEquationOfElectricalCircuitsE x

56
4.3 Analog Filters
Several filters can be easily implemetned uisng analog filters such as a low pass filter, high
pass filter, and lead/lag compensator, etc..Therefore, it is more convenient to use electic
systems rather than mechanical systems. In below, some analog filters are introduced
with their transfer functions of ouput/input represented in s domain.
- Low pass filter
Vin = C1 idt + Ri Cs 1
R 
+R I
Vo = C1 idt = Cs I
R
I
Vo I
Vin
= 1
Cs
+R
= 1+RCs
Cs

The low pass filter can be used for elliminating the electrical noises occured in a
high frequency because of its reducing gain after the cutoff freqeuncy. It is also used for
integrating purpose. For example, when the displacement needs to be obtained from the
velocity, the low pass filter can be used. Of course, an inegrator 1s can be used, however,
the DC or offset values make the signal drift to saturate.Hence, the low pass filter can
be used instead of the integrator for the high frequncy signal.

Q: How is the cutoff frequency of the low pass filter determined to obtain
the displacement from the velocity signal sin(100Hzt).

- High pass filter

Vo RCs
Vin
= 1+RCs

- Lead compensator/ Lag compensator

- Lead lag compensator

57
4.4 The Load effect modeled using an electronic circuit
A load effect is easily understood using an electronic circuit model in physical systems.
When a load is modeled using a resistor, the, current is obtained form the electical
relation of i = Re as shown in Fig. 8.2.

Figure 3.33: A load modeled using a resistor

When another load is added in series or pararell to the inital circuit as shown in Fig.
8.3, it is unfortunate not to use the analysis done for the initial curcuit. A new analysis
should be done for a new cicuit to calculate an effort or a flow variables because the
system is passive.

58
Figure 3.34: An electric cicuit with another load added in parallel

Similary, this load effect is also found in a mechanical system. When a load is applied
to the initial load in series as shown in Fig.8.4, a new analysis for the 2 DOF mechanical
system should be done even though we finish the initial load analysis.

Figure 3.35: Mechanical system with another load added in series

For example, when an additional components are added to the initial circuit com-
posed of R1 and C as shown in Fig. 8.1, the electrical equations associated with the
circuit are represented as

59
Z
1 L{v(t)} 1
1) v (t) = R1 i (t) + i1 (t)dt −−−−→ V (s) = R1 I (s) + I1 (s)
C Cs
L{i(t)}
2) i (t) = i1 (t) + i2 (t) −−−−→ I (s) = I1 (s) + I2 (s)
Z
1 L{vo (t)} 1
3) vo (t) = i1 (t)dt = R2 i2 (t) −−−−−→ Vo (s) = I1 (s) = R2 I2 (s)
C Cs
Vo
Then, the transfer fucntion of output/input V
is obtained as
1
Vo I (s)
Cs 1
= 1
V R1 I (s) + Cs I1 (s)
1
I
Cs 1
(s)
= 1
R1 [I1 (s) + I2 (s)] + I
Cs 1
(s)
1
= h i
R1
R1 Cs + 1 + R2

The transfer function Vo /V , is not only dependent on R1 , C but also dependent of R2 ,


which indicates a load effect occurs due to the another load R2 . In other words, the intial
circuit is not isolated form the second circuit. It is because they are passive circuits.
In conclusion, it is unavoidable to have a non-isolated system if the system is passive.
In order to have an isolated system, a system should be feedback system(servo system)
such as an op-amp.

4.5 Operational amplifier free from a load effect


An operational amplifier is one of typical examples which takes an advantage of no
load effect using an electronic feedback system. It has characteristics of a high input
impedance and low output impedance.

The impedanceof a device Z is defined as ei . Let’s look at a simple electronic circuit


composed of a resistor R and capacitor C shown in Fig. 8.6(a). 8.6(b) shows its bond
graph model showing all components and 8.6(c) shows an equivalent bond graph modeled
using just an equivalent resistor Req . The equation of the electronic circuit is represented
Z
1 1
e = Ri + idt = (R + )I (3.21)
C Cs
Hence, the impedance of the resitor Z is is obtained as
e 1
Z= =R+ (3.22)
i Cs
60
Figure 3.36: A simple electronic circuit composed of a resistor R and capacitorC

As represented in Eq.8.3, Z is depedndent on frequency,

61
4.5.1 Voltage amplifier

Figure 3.37: An operational voltage amplifier

Figure 8.8 shows an operational voltage amplifier where the output voltage is propor-
tionally amplified to the input voltage. Due to the condition of high input impedance of
the op amplifier, the input current iin can not flow through the resistor but flow through
the feedback resistor Rf . In other words, the feedback loop makes the output current if
equal the input cuurent iin . These conditions lead to the following resuts.

v+ = v− = 0 (3.23)
vi − v − vi
iin = = (3.24)
R1 R1

v − vo vo
if = =− (3.25)
Rf Rf
iin = if (3.26)

From the above equations, we have

Rf
vo = − vin (3.27)
R1
The name of the voltage amplifer comes form the concept of the feedback of the
output voltage vo as the name of the position controller(amplifier) comse frome the
postion feedback(vo is equal to -vf ). This relation is proven as below
v1 Rf
vf = Rf if = Rf = v1 = −vo (3.28)
R1 R1
Recalling the ideal condition of the voltage source in section 3.2 of Chapter 1, the op
amplifier is consdiered as a voltage source because it has a low output impedance, that

62
is, no voltage loss from an internal resistor. Therefore, when a load RL is connected to
the ouput voltage, the current is determined from the relation of RVoL and it is drawn from
a power supply. Here, RL is supposed to be much larger than the output impedance of
the op amplifier.

It is possible to make several isolating systems using op amplifers having unit gains(or
buffers) between the passive systems as shown in Fig. 8.9. Hence, there is no need to
do circuit anaysis again when a new passive circuit is added to the intial circuit. It is a
great advantage of using operatiomal amplifiers.

Figure 3.38: Op amplifers between the passive systems to make isolating systems

4.5.2 Current amplifier

Figure 3.39: An operational current amplifier

Figure 8.12 shows a current amplifier used for amplifying the current proportionally to
the input voltage. There is a feedback loop connecting the point having Va to the point
having v . Due to the condition of high input impedance of the op amplifier, the input
current iin can not flow through the resistor of the op amplifier but flow through the

63
feedback resistor Rf . In other words, the feedback loop makes the output current if
equal the input cuurent iin . These conditions lead to the following resuts. The load
current IL is divided into the feedback current if and Is . The above conditions lead to
the following results

v+ = v− = 0 (3.29)
vi − v − vin
iin = = (3.30)
R1 R1
vs
if = = −iin (3.31)
Rf
 
vi Rf
vs = Rf if = Rf − = − vi (3.32)
R1 R1
 
vs 1 1
iL = if + is = + vs Rs = vs + vi (3.33)
Rf Rf Rs
 
vin Rf
vs = Rf if = Rf − = − vin (3.34)
R1 R1

From the above equations, we have the relation that the ouput current IL is amplified
by the input voltage vin as
 
Rf 1 1 Rf (Rs + Rf ) Rs + Rf
iL = − + vin = − vin = − vi (3.35)
R1 Rf Rs R1 Rf Rs R1 Rs

The name of the current amplifer comes form the concept of the feedback of the
output current iL though the actual feedback current is the partial of it. Recalling the
ideal condition of the current source in section 3.4 of Chapter 1, the current amplifier is
consdiered as a current source because it has a high input impedance, that is, no current
loss from an internal resistor. Therefore, when a load RL is connected to the ouput,
the current is determined by the input voltage regardless of the magnitude of the load
resistor and it is drawn from a power supply. Here, RL is supposed to be smaller than
the input impedance of the op amplifier.

Rs is used to flow the current to the direcion to the ground because Rf is usually
chosen to be much larger than RL . Hence, Is ∼= IL . If Rs does not exist, the high current
output will flow to Rf , which will burn out the op amplifier. Rs is also used to monitor
IL in terms of voltage. The power consumption in Rs is calculated as Is2 Rs . So, the
material property of Rs should be strong enough not to be burnt.

64
5 Electronic circuit model of a pneumatic system
Let us think about a pneumatic system shown in Fig. 3.40. It is composed of a com-
pressor, on-off valve, and nozzle. It is used for blowing various metal scraps such as
Al, Cu, Pb, ans steel to containers located at different distances from the nozzles for
classifying different materals of metals by applying appropriate pneumatic flow rates
to metal scraps. When the on-off valve is open, the maximum pressure is built up at
the compressor with zero flow rate. When the on-off valve is closed, the pressure newly
determined by the load supplies the flow rate at the nozzle and it blows a specific metal
scrap to fly to a container.

Figure 3.40: Configuration of a pneumatic system

An electronic circuit model is used for easy undstanding of the pneumatic system
and is shown in Fig. 3.41(a) when one nozzle is used. The on-off valve is modeled using
a switch. The nozzle is modeled using a resistor Rn . The compressor pressure Pc is
maiximum when there is no flow rate. The bond graph representation of this pneumatic
system is shown in Fig. 3.41(b). Here, Pc ∗ Qc = Rn ∗ Q2 + P2 ∗ Q(Qc = Q). When
P2 = 0, we have P c = Rn Q. Hence, we can easily find out the flow Q graphically using
the load curve and the relation of P c ∗ Qc = const as shown in Fig. 3.42.

65
Figure 3.41: (a) An electronic circuit model of the pneumatic system having one nozzle
and (b) its bond graph representation

Figure 3.42: The load curve of the pneumatic system

When the flow Q determined from the load is large, then the metal scrap will fly far-
ther than the desiganted distance. Of course, this problem can be solved using a larger
Rn . However, it is not practically possible to obtain a nozzle which has a proper Rn
becasue of supply difficulty of various types of nozzles in a market. Instead, a regulator
can be used to adjust overall resistance in the pneumatic system. Its schematic diagram
is showin Fig. 3.43 The regulator in the pneumatic system is modeled using a resistor
Rreg and a capacitor Creg as showin in Fig. 3.44. The regulator is used to have a desired
pressure P1 and flow rate Q for operation.

66
Figure 3.43: The schematic diagram of a regulator

Figure 3.44: An electronic circuit model of the pneumatic system having a regulator

The bond graph of the pneumatic system with the regulator is represented in Fig.
3.45. Rreg and Creg decrease Pc to P1 becasue the pressure Pc keeps decreassing through
the regulator loads.

67
Figure 3.45: The bond graph of the pneumatic system when a regulator is used.

The equations associated with the bond graph showin in Fig.3.45 are
Pc − P1
Qc = (3.36)
Rreg
Z
1
P1 = (Qc − Q1 )dt (3.37)
Creg
P1 − P2
Q = (3.38)
Rn
When P2 = 0, dt = 1sec, Rn = 1(unit), Q is obtained as

Pc
Q= (3.39)
Rreg Creg + Rreg + 1

As expressed in Eq.3.39, a proper flow rate Q to blow a desired distance is determined


by Pc , Rreg , and Creg .

Q: How is the load curve changed when a regulator is used? Show how the
flow rate Q is reduced.

It is required to use multiple nozzles when several scraps need to blow at the same
time modeled using an electronic circuit as shown in Fig. 3.46. When the same resis-
tance as the single nozzle Rn is used for indivisual nozzles, the equivalent resistance of
the multiple nozzle decreases because they are considered as the resisors connected in
parallel. Hence, a higher flow rate is required to distribute to multiple nozzles. Hence, a
Rreg , and Creg are properly adjusted to supply a higher flow rate to the multiple nozzles.
If Pc and Qc decrease so much, it is not possible to flow a sufficient flow rate to the
multiple nozzles.

68
Figure 3.46: An electronic circuit model of the pneumatic system when multiple nozzles
are used

Q: Draw a bond graph of the pneumatic system modeled using an electronic


circuit shown in Fig.3.46. Explain how the regulator is designed using a load
curve.

Figure 3.47: Various conditions of pressure vs flow rate when a regulator is used or not
used and single nozzle or multiple nozzles are used.

6 Hydraulic system
Draw a bond graph of a hydraulic system shown in Fig. 3.48. Drive its state equations.
Drive a transfer function of the hydraulic system when input is Qand output is Q2

69
Figure 3.48: A hydraulic system composed of tank and valve

We can also obtain a transfer function from the state equation

If Q2 > Q, v1 is reduced, hence P1 is reduced. This result explains why P1 is reduced


when Q is not much supplied in a pump(or compressor). The 0-junction represents the
practical power source in which pressure drop occurs when flow is not constant(or flow
or increased).

Initially, when there is one valve R2 , the flow is assumed to be Q1 . When other valves
connected to R1 the flow is increased to 3Q, because the effective resistance is reduced
to R2 /3 . R1 can be experimentally obtained by recording P versus Q1
Here, P1 is decreased because Q1 is increased to 3Q1 . P is increased because the flow is
increased to 3Q1 , then P2 is decreased.

For example, P1 = 7bar, P = 2bar, P2 = 5bar.

70
When the flow is Q1 , then P1 = 5bar, P = 3bar, P2 = 2bar for 3Q1 .

The transfer function QQ2 for the system with one valve can be obtained from the state
equation. The state equations are

 
˙ 1 1 V1 V2
V1 = Q − Q1 = Q − (P1 − P2 ) = Q − −
R1 R1 C1 C2
 
˙ 1 P2 V1 1 1 1
V2 = Q1 − Q2 = Q − (P1 − P2 ) − = − + V2
R1 R2 R1 C1 R1 R1 C2 R2 C2
Q2 P2 V2
Then, since Q
= QR2
= QC2 R2
, from the state equations

SV1 = Q − Q1
SV2 = Q1 − Q2

We can obtain

Q2 1
Q
= (s+1)(s+2)−1

for the simple case of R1 = 1, R2 = 1, C1 = 1, R2 = 1.

71
7 Ideal 2 port element: T and G
There are an ideal 2 port element, a transformer T and gyrator G.

Figure 3.49: Examples of T element

It has derivative causality problem, so it should be changed using a spring between the
pulley and the mass 1

72
73
7.1 Coupling element causality

Once the effort on one bond is determined


then the effort on the remaining bond is
specified.

74
8 Some other electronic circuits
- Charging-discharging circuit
Figure 3.50 shows a charging and discharging circuit using a diode, capacitor, and resis-
tor. Let’s find out how the transfer fucntion VVoi is expressed in the frequency domain.

Figure 3.50: Charging and discharging circuit

A diode characteristic V − Iis represented in Fig. ?? showing it is just a nonlinear


resistor whose resistance varies for the magnitude of the applied voltage. When a for-
ward voltage is applied to a diode, the current does not flow until the applied voltage
reach to some voltage level 0.5 V . When a reverse voltage is applied to a diode, the
current does not flow because it plays as a high resistor.

A bond graph model or a electronic circuit model can be used to obtain governing
equations of the electronic circuit. In a charging cycle, namely Vi ≤ Vo , we have electronic
circuit equattions as

Vi − RD i = Vo (3.40)
1
Vo = i2 RL = i1 (3.41)
Cs
i = i1 + i2 (3.42)
Vo
Then, the transfer fucntion Vi
is obtained as

Vo 1
= (3.43)
Vi 1 + RD Cs
In a discharging cycle, namely Vo ≤ Vi , the current charged in the capacitor C flows
to only RL because of the reverse diode. The electronic circuit equattion is represented

75
as
Z
1
idt + RL i = 0Vi − RD i = Vo (3.44)
c
1
Vo = i2 RL = i1 (3.45)
Cs
i = i1 + i2 (3.46)

the transfer fucntion VVoi isrepresentedas VVoi = 1


1+RD Cs
f inalW henapulseinputisapplied
Optical switch

76
Chapter 4

Electromechanical system

1 magnetic system

77
Chapter 5

Signal Processing

Engineers are usually required to design a sensor that is suitable to the application pur-
poses. There are important specifications we need to consider for sensor deign, those
are speed, accuracy, and resolution(precision). The definitions of three specifications are
described in class.

1 Signal to noise ratio of a sensing system


Q: For an AFM system to have resolution of 0.5nm, how good precision a
sensor should have for a maximum measurement distance of dmax = 1µm.

dmax : Vmax = dmin : Vmin

78
Vmin Vmin
dmin = dmax = × 1µm = 0.5nm
Vmax Vmax

Vmin 1
= → resolution is 2000
Vmax 2000

(sol) So, when Vmax = 10V , Vmin = 5mV , the sensor noise should be less than 5 mV.

It is important to design a sensor with small electronic noise. Modulation and demodu-
lation techniques can be used for improving sensor precision even though they have been
applied to other industrial areas such as information and communication and electron-
ics. Some typical signal processing techniques are introduced with an emphasis on real
applications to sensing systems.

2 Optical sensor modulation


A laser diode is one of popular components as a light source for an optical sensor. Current
should supply to the laser diode to generate light. Figure 5.1 shows an electronic circuit
used for current amplication using a transisitor, combined name of tranfer and resistor.

79
Figure 5.1: (a) Transistor for current amplification and (b) real configuration of transi-
tors

There are three legs, called a base, emitter, and collector in a transistor. It controls
the amount of current that flows from the collector to the emittor depending on the
amount of current flowing into the base. The elcctronic circuit equations associated
wuth Fig. 5.1 are derived in

VBB = Rb ib + VBE (5.1)


ic = βib (5.2)
ie = Ic + Ib = (β + 1)ib (5.3)
VCC = Rc ic + VCE (5.4)

Here, VBB is the supply voltage in the base. VCC is the supply voltage in the collector.
β is the current amplifying gain, that is 200 ∼ 300. VBE is the voltage drop for its
operation between the base and emitter and it is 0.5 ∼ 0.7V. VCE is the voltage drop
between the collector and emitter. From the above equations, we have

VBB − VBE
ib = (5.5)
Rb
When ib is determined from Eq. (5.5), the current ic is amplifed regardless of VCC .
Therefore, it can be said that the transisitor can be used for a current amplifier. Prac-
tically, as the characteristic curve of the transistor shown in Figure 5.2, ic increases
proportionally to VCE at the voltage less than 0.7 V and it is saturated to a constant

80
current at the voltage larger than 0.7 V . As shown in Fig. 5.2, there is no big constraint
on VCE for a good operation of the transisor since almost constant current can be ob-
tained regardless of its mangitude when the proper range of Rc is selected. It is more
important to decide ib that is determined by VBB and Rb .

Figure 5.2: The characteristic curve of the transistor

For example, if you want to run 100 mA on a motor or laser diode, let’s see how
much resistance is needed in Rb . First, assuming the current gain β is around 100, the
current that should flow to the base ib is 1 mA from Eq.(5.5). When Vb is 5V , and Vbe
is 0.7V , the voltage across Rb is 4.3V . Then, Rb = 4.3V /0.001A according to Eq. (5.2),
so you can attach 4.3 kΩ.

We can find out a mechancal device that is equivalent to the transistor. Figure
5.3 shows a servo valve used in a hydraulic system. When the spool operated by the
mangetic force is slightly away from the netural position, the flow increases lineary with
respect to the opened area of the valve and enters into the cylinder to move the piston.
The spool can move using a very small force as a very low current in the base is required
to produce the large current in the collector.

81
Figure 5.3: Transistor operating eqiavalently as a servo valve

Q: When 10V VCC is used, select an appropriate resistance Rc to make the


transistor shown in Fig. 5.1 work well for current amplification. Measure
VCE using an oscilloscope.

3 Amplitude modulation
We can increase the precision (or S/N) of a sensing system using modulating techniques.
One candidate is amplitude modulation by shifting the driving frequency of the light
source to somewhere external noises are minimum. The frequency of the position will
be shifted to both side bands of the driving frequency ωc . The irradiance I(t) of the
sinusoidal modulated light source can be expressed as

I(t) = d + m sin(ωc t) (d > m)

82
which is shown in Fig. 5.4. where d and m are constants and denoted as DC offset and
the amplitude of sinusoidally driven signal at frequency ωc .

Figure 5.4: Sinusoidally modulated irradiance of the light source for a PSD sensor

The DC offset d is applied because irradiance cannot be negative. To prevent the


modulated irradiance from be negative or being discontinuous, d must be larger than m.
When the external noise whose irradiance ζ(t) exists, the current measured at a PSD
sensor is expressed as
1
i(t) = K[d + m sin(ωc t + α) + ζ(t)] = K[d + m sin(ωc t + α) + ζ(t)]X(t)
x(t)
Here, K is a proportional constant, α is the phase lag due to the propagation speed in

83
the PSD and the processing circuit. x(t) is the position of light source. ωc is a carrier
frequency. To cancel out the current induced by external noise and DC offset of the
light source, the signal i(t) is feed to a band pass filter centered at ωc , then the currents
induced by external noises and DC offset of light source can be eliminated. Therefore,
the driving frequency of the light source is designed to have far distant from the noise
frequency to make it easy to cancel out the noise by using a BPF. Then, the resulted
signal contains only the current induced by the sinusoidally driven light source. That is,

i(t) = Km sin(ωc t + α)X(t)

In conclusion, the above equation indicates that the frequency of the motion of X(t) is
shifted by the carrier frequency ωc . The shifted frequencies are ωc − ωm and ωc + ωm
where external noises of X(t) are minimum.

Q: Draw the above equation when X(t) is slowly moving with sinusoidal
motion sinωm t. When ωm = 1 × 2πHz, ωc = 50 × 2πHz

Q: What are the results of a LPF and a HPF of Km sin(ωc t + α)X(t), with
sinusoidal motion sin ωm t.

84
Physical meaning of modulation

If we interpret the amplitude modulation process in the frequency domain, i(t) can be
rewritten with the assumption α = 0,X(t) = sin ωm t for simplicity as
1
i(t) = Km sin(ωc t + α) sin ωm t = [cos(ωc + ωm )t − cos(ωc − ωm )t]
2
i(t) in the frequency domain indicates that the frequency of X(t) is shifted to ωc + ωm ,
ωc − ωm . The frequency of the position signal is shifted to both side bands of the driving
frequency ωc . We can take an advantage of shifting the frequency of motion to some-
where external noises are minimized.

Amplitude demodulation

In order to obtain X(t), demodulation process is required. If we multiply sinωc t with


i(t),

i(t)(sin ωc t) = Km sin ωc t sin(ωc t + α)X(t)


Km
= [cos α − cos(2ωc t + α)]X(t)
2
The carrier frequency ωc can be eliminated by applying a LPF to the above signal, that
is,
Km
V1 (t) = LP F [i(t)(sin ωc t] = cos α X(t)
2

85
At the same way, i(t) multiplied by cosc t can be obtained as represented

i(t)(cos ωc t) = Km cos ωc t sin(ωc t + α)X(t)


Km
= [sin α − sin(2ωc t + α)]X(t)
2
The carrier frequency ωc can be eliminated by applying a LPF to the above signal, that
is,
Km
V2 (t) = LP F [i(t)(cos ωc t] = sin α X(t)
2
To eliminate the dependence on α from the above equations, we introduce a trigonomet-
ric relation, sin2 α + cos2 α = 1.
Km
X(t)]2
[V (t)]2 = [V1 (t)]2 + [V2 (t)]2 = [
2
The above eqn. reveals that the resulted signal contains only X(t) induced by the given
light source. No dependence in the phase delay is concerned in the signal. The current
i(t) induced by only the light source given can be obtained by calculating square root
of the above eqn. as
r
Km Km 2
V (t) = [ X(t)]2 = X(t)] T heref ore, X(t) = V (t)
2 2 Km
Simply, the modulated signal i(t) = Kmsin(ωc t + α)X(t) can be demodulated using
another method, RMS(root mean square) circuit as by squaring i(t) first,

[V (t)]2 = [i(t)]2 = (KmX(t))2 sin2 (ωc t + α)


(1 − cos 2(ωc t + α)
= (KmX(t))2
2
(KmX(t))2
Secondly, the mean of the signal [V (t)]2 is 2
.

Q: Why is the averaging process of V (t) approximately equal to LPF of


V (t)?


X 2πkt 2πkt
V (t) = a0 + (ak sin + bk cos )
m=0
T T
Z T
1 2 1
a0 = V (t)dt = (Ṽ (t)T ) = Ṽ (t)
T −T
2
T
LP F [V (t)] = a0 = Ṽ (t)

Km 2
Finally, the root of the signal V (t) is √ X(t).
2
Hence, X(t) = Km
V (t).

86
4 Amplitude modulation applications in a real sys-
tem(AFM)
AFM (Atomic Force Microscope) is a device which measures the properties of sample
surface using the physical force between probe tip and sample surface. AFM is sorted
out contact type and non-contact type.

The configuration of an AFM

A real AFM configuration developed at Em4sys. Co. Ltd. is as follows

87
The principle of the contact AFM

When a tip is in contact with a sample surface, a repulsive force generates in the can-
tilever. The repulsive force bends the cantilever whose angle is measured from an optical
sensor mounted in the AFM head. The bent angle is proportional to the sample height.
The motor motion to make the bent angle the same as before. If the repulsive force
or bent angle are controlled to be constant, the sample height can be obtained by the
moving displacement of the motor from the statement addressed in below. Refer to
https://www.ntmdt-si.com/resources/spm-principles to understand its working princi-
ple with video demonstration.

88
C = zdist1 + ∆z + z1 = zdist2 + ∆z + z2
zdist1 = C − (∆z + z1 )
zdist2 = C − (∆z + z2 )
The heights of the sample dist1 and dist2 at different locations are obtained as dist1 =
C − (∆z + z1 ) and dist2 = C − (∆z + z2 ), respectively. If ∆z and z are measured, dist are
easily obtained. However, it is difficult to mount two sensors on a small space with the
cantilever and the motor, we design a AFM head using just one sensor that measures
z. Furthermore, if ∆z is controlled to be constant using force control as shown in Fig.
dist are obtained by z from Eq. . Since there is no sensor for z measurement we
are looking for a possibility if we can obtain dist using some variables inside
the feedback loop. Fortunately, there seems to be some relation between z and the
controlled output. We just pick up a controlled output directly from the force control
loop because the controlled output is proportional to the plant output z under the as-
sumption that there is no dynamics in the amplifier, actuator, and plant.

The principle of the noncontact AFM

When there are abrupt changes of height in sample, the fragile cantilever can be easily
broken. To avoid its failure due to mechanical damage, a noncontact type cantilever is

89
used. It is mechanically vibrated(modulated) close to a sample surface in the air with
the motion of sin ωm t. Then, an attractive atomic force varies with the distance between
the tip and surface.

Figure 5.5: A block diagram of noncontack AFM

Suppose that a cantilever has vibrational motion when it is excited by a PZT actuator
driven at frequency fex close to its natural frequency at a gap 1. The amplitude of
the vibration M1 is determined from the frequency response function. When the gap
increases it natural frequency decreases because the spring stiffness decreases. When
the gap decreases it natural frequency increases because the spring stiffness increases.
Hence, the changed frequency makes the amplitude of the cantilever changes. When a
driving voltage of a PZT actuator is described as

V (t) = V sin 2πfex

Then, a cantilever is modulated as

Z(t) = S(t) sin(2πfex + θ)

90
Here, S(t) is dependent on the different gap magnitude, which is described due to the
statement addressed in the above. Fig. shows how Note that Eq. is the same expression
as in electronic amplitude modulation. In another word, the amplitude modulation can
be realized mechanically to take its advantage of noise reduction. The next step is to
obtain S(t) by using a demodulation technique.

Equivalence of attractive force variation in the air to spring stiff-


ness variation

Q: What is an advantage of using a modulated technique of a cantilever


compared to a method without modulation?

We take the similar procedure of demodulation technique as we did in the previous

91
work. If Z(t) is multiplied with V (t)

V1 (t)Z(t) = V S(t) sin(2πfex t) sin(2πfex t + θ)


1
= − V S(t)[cos(4πfex t + θ) − cos(−θ)]
2
1
= V S(t)[cos(4πfex t + θ) + cos θ]
2
The carrier frequency fex can be eliminated by applying a LPF to the above signal,
Then, we have
1
LP F [V1 (t)Z(t)] = LP F [i(t)(sin ωc t)] = V S(t) cos θ
2

At the same way, when Z(t) multiplied by 90 phase different signal is V2 (t)Z(t) then,

V2 (t)Z(t) = V S(t) cos(2πfex t) sin(2πfex t + θ)


1
= − V S(t)[sin(4πfex t + θ) − sin(−θ)]
2
1
= V S(t)[sin(4πfex t + θ) + sin θ]
2
By applying a LPF to the above signal, we have
1
V2 (t)Z(t) = V S(t) sin θ
2
p
We can obtain S(t) from solving (1)2 + (2)2 equation. Alternatively, we can also ob-
tain S(t) using a RMS circuit as used in Section 2.4.
Figure show a block diagram for force control of a noncontact type AFM.

Q: What modulation signal can we get when a low frequency vibration


technique get rid of When a cantilever is modulated using a PZT actuator
driven by the signal with a frequency of fex .

92
Photo detector circuit

When a laser beam falls on a photo diode, photocurrent is generated in the photo diode.
The current is amplified using a trans-impedance amplifier circuit. TIA is a current to
voltage converter, almost exclusively implemented with one or more operational ampli-
fiers.

Vo ut
Q: Derive the transfer functions of the above TIA circuits Ii n
. Design a
TIA circuit which can realize your purpose.

93
Q: Investigate all TIA circuits used in Sensor and Actuator Lab. Draw
electronic circuits using the real values of electronic components and derive
their transfer functions. Explain their design criterion.

5 Phase modulation
In the AFM addressed in section 5, when a driving voltage of a PZT actuator is described
as
V (t) = V sin 2πfex
a cantilever is modulated as

Z(t) = S(t) sin(2πfex + θ)

θ(t) is also changed as well S(t) depending on gap difference. In other words, the phase
is modulated as S(t) is modulated. θ(t) is demodulated using the method introduced in
section 5.
Another application is found in the light interference phenomena.

interference

Wave equation
2π 2π
E(x, t) = E0 ei( λ x+ T t+φ)

94
Here, k = 2πλ
is spatial wave number that is equivalent to ω = 2π
T
.λ is a period in
displacement as T is a period in time. E(x, t) = E0 ei(kz+ωt+φ) .

Static interference is observed using two light source or laser beams and it is represented
using a sinusoidal equation.

Dynamic interference is generated when there is motion as in a simple interferometer

95
The two laser beams interfered in the photo detector is represented as in Eq.

E = |E1 + E2 |2 = (E1 + E2 ) × (E1 + E2 )∗


= |E1 ei(kx1 +2πf1 t+φ1 ) + E2 ei(kx2 +2πf2 t+φ2 )2 |
= (E1 ei1 + E2 ei2 ) × (E1 e−i1 + E2 e−i2 )
= E12 + E22 + E1 E2 ei(1 −2 ) + E1 E2 e−i(1 −2 )
= E12 + E22 + 2E1 E2 cos(1 − 2 )

if E1 = E2 = E0

E = 2E02 + 2E02 cos(k(x1 − x2 ) + 2π(f1 − f2 )t + (φ1 − φ2 ))


= 2E02 (1 + cos(k + 2π∆f t))

where ∆x = x1 − x2 , ∆f = f1 − f2

If ∆f = 0, it is called a homodyne interferometer. Since ∆x = 2x,E(x, t) =


2E02 (1+ cos 2kx). Note that x is modulated in the phase. Hence, it can be said that an
interferometer is one of phase modulators.

96
Q: How is the interference signal E(x, t) represented for ∆x = λ2 , λ and 0 re-
spectively?

For ∆x = λ2 ,E = 0, for ∆x = λ,E = E02 , for ∆x = 0, E = 4E02 .

In general, when there is ∆f , it is called a heterodyne interferometer. A typical example


of the optical configuration of a heterodyne interferometer is shown in Fig. 5.6.

Figure 5.6: The configuration of a heterodyne interferometer

97
The interference signal is represented using Eq.

E(x, t) = 2E02 [1 + cos(2π∆f t + ∆x)]
λ
∆x
= 2E02 [1 + cos 2π(∆f t + )]
λ
2x
= 2E02 [1 + cos 2π(∆f t + )]
λ
2x
= 2E02 [1 + cos 2πt(∆f + )]
λt

Here, 2x
λt
is called a Doppler frequency that is obtained by differentiating the phase
term(∆f t + 2xλ
).
2x 2v
fd = = (5.6)
λt λ
Eq.(5.6) indicates that velocity of motion is proportionally obtained by Doppler fre-
quency. Therefore, it can be said that velocity is obtained by a frequency demodulation
technique.

Q: Sketch the interference signal processed with a HPF whose cutoff fre-
quency 10Hz when we have displacement motion of cos 2(100Hz)t.

98
Q: What is the physical meaning of Eq?

Q: Why the velocity signal is preferred to use compared displacement signal?

99
LIDAR(Light detection anc ranging) AMCW sensor

Another application example of phase modulation is found in LIDAR(light detection


and ranging) sensor. LiDARs (light detection and ranging), also known as LADARs
(laser detection and ranging), are widely used in numerous engineering fields, such as
autonomous vehicles, robot navigation, and three-dimensional scanning applications,
because of their long-distance high-precision measuring capabilities in outdoor environ-
ments [1], [2]. Avalanche photodetectors (APD) are commonly used for LiDARs, as
their high internal gain can considerably amplify a measured signal.
There are several methods in Lidar in terms of principle. A pulsed time-of-flight method
is to measure the distance to an object by measuring the time difference between the
reference and measured signal, as shown in Fig. 5.7. When the time difference is ∆t,
the distance to the object, d, is obtained as

∆t
d=c×
2
where c is the speed of light, and ∆t is divided by two as the signal travels in two direc-
tions [11].

Figure 5.7: Measurement principle of pulsed time-of-flight method

In the AMCW method, the distance that light travels can be determined using the
phase difference between the modulated reference signal and measured signals, as shown
in Fig. 5.8. When light travels a distance d, light reflected from an object is delayed
by the phase, ϕ. The phase difference between the reference and measured signals is
proportional to the distance traveled. One thing to note here is the delayed phases are
the same regardless of the modulating reference frequencies. From Fig. we have the
relation of
c = f λc = fm λm

100
where c(= 3 × 108 m/s) is the speed of light, fm is the modulating frequency, λm is the
wave length of modulating signal.
2D ϕ 4πD
= → ϕ=
λ 2π λ

When the modulated frequency of the reference signal is fm , the distance D is ob-
tained in terms of phase ϕ and fm as
ϕλm cϕ
D= =
4π 4πfm
Therefore,
ϕm c δϕ
δD = =
4π 4πfm
Hence, with a higher modulated frequency, a better distance resolution can be obtained.
Mathematical expression of the modulation signal, Im (t) is

Im (t) = Am cos(fm t + ϕ)

If we find the phase difference ϕ that is proportional to the distance using a demod-
ulation technique, the traveled distance is obtained. We can apply the demodulation
technique introduced in the above. Mixing the reference signal Vr (t) = Ar cos(fm t) with
the modulation signal, Im (t) gives Vo (t)

V0 (t) = Ar Am cos(fm t)cos(fm t + ϕ)


1
= Am Ar [cos(2fm t + ϕ) + cosϕ]
2
A low pass filtered output gives the phase as a result. This kind of processing is also
called a Lock-in amplifier.

101
Figure 5.8: System configuration of the AMCW lidar

Alternatively, we can find the phase by calculating the overlapped area of the rectified
signals.

102
Q: Suggest a method how the phase is calculated.

-Find out a mean value of the signal.


-Take a low pass filter processing of the signal.
-Find out a cut-off frequency of the LPF which makes the above results are the same.

c δϕ
From the relation of δD = 4πf m
, the phase difference is the same for the same dis-
tance difference regardless of the fm . Hence, the frequency shift technique is applied to
have a higher resolution.

If a modulated frequency 40MHz is applied, the detecting signal Im (t) with noise N (t)
is obtained in a detector as

Im (t) = Am cos(40M Hz t + ϕ) + N (t)

Mixing Im (t) with 39.95MHz signal, then we obtain the intermediate output Ii (t) as

Ii (t) = Am Ar cos(39.95M Hz t)cos(40M Hz t + ϕ) + N (t)


1
= Am Ar [cos(79.95M Hz t + ϕ) + cos(0kHz t + ϕ)] + Am Ar cos(39.95M Hz t)N (t)
2
Then, low pass filtered Ii (t) is 21 Am Ar cos(50kHz t + ϕ) that is much clean signal com-
pared with Im (t) as shown in Fig.

103
6 Frequency modulation
The general equation for describing modulation techniques is

Im (t) = Am cos(fm (t)t + ϕ(t))

When the information to be found is A(t), we call it amplitude modulation. When the
information to be found is ϕ(t), we call it phase modulation. Lastly, the frequency fm (t)
can be also modulated as you expect. It is found in FMCW LiDAR.

104
Figure 5.9: Block diagram for self-chirped homodyne detection

105
Frequency modulation is achieved by using an EOM that is represented in below.

106
Chapter 6

Control

1 Introduction
It is best to make a good design of a plant that can perform its tasks without additional
compensator that is usually called a controller or filter. Therefore, we need to put our all
efforts as much as possible in having good plant characteristics. However, it is not easy
to design a plant that realizes its performances as expected. Sometimes, it has unstable
factors which cause a large steady state error. Additionally, it has a very small damping
ratio that cause a large vibration in motion. Moreover, it has uncertainty in a high
frequency region which makes it difficult to model correctly. These kind of problems can
be solved by using a controller. However, it is also important to recognize a feedback
system does not always provide satisfactory results in terms of performances such as
steady state error, stability, and speed.
There are two analysis methods in controller design process, that is a closed loop
analysis(time response analysis) and open loop analysis(frequency response analysis).
First, we begin with a closed loop analysis. It has advantages of straightforward process
and easiness of understanding of a feedback system. However, it is very complicated
and tedious repeating job whenever the order of a plant P (s) is increased and controller
C (s) changes. In this analysis, the controller can not be seperated from the plant. Next,
we will move to an open loop analysis showing its advantages of simple process in design
a controller using a plant P (s) only.

107
2 Why is a feedback loop needed?
If an input is applied to a plant, the output response is decided based on its mathematical
model. We can expect how the output motion would be from the known model. How-
ever, the real system can be different from the mathematical model, which is definitely
occurred in a high frequency. Therefore, the output response would be different from
what we expect. For example, when a unit input is applied to an object that modeled
as mass-spring-damper system, we know how the displacement of the object would be
in a steady state if their parameters are the exactly same as those of the matematical
model. However, if there are some difference between them in reality, the displacement
of the object is different from the result that is obtained from the mathematical model.

Figure 6.1 shows a typical example of position feedback system composed of an


amplifier, actuator, plant, and sensor. The amplifier and sensor are usually made of
electrical devices becuase they are easily built up whose inputs and outputs are voltage
levels. The actuator converts the electrical variables such as voltage or current to me-
chanical power variables such as force or velocity. The plant converts the mechanical
effort vairables such as force or torque to flow variables such as velocity or displacement.
If the command or reference input is position, we need a position sensor.

Figure 6.1: The position feedback system

When the command signal applies, the acuator supplies force to the plant and it
moves close to a target position. Then, the sensor signal increases, which makes the
error signal decrease. The acuator supplies more force to the plant and it moves closer
to a target position. Finally, the sensor signal is the same as the command signal re-
sulting in a zero error. When the sensor signal is largen than the command signal, the
negative error supplies a negative force that drives the plant in a revese direction and
finally reach to a target position.

Though the plant is not modeled well, in other words, the system has a modeling

108
uncertainties in parameter estimation or system structure, the sensor signal reflects the
current situations. Therefore, the error signal can compensate the plant so that it can
follow the command input.

3 Feedforward transfer function


The feedforward transfer function G(s) includes the amplfier, actuator, plant, and sensor.
It is very important to know the dynamics of the components of the amplifier and
sensor before a controller is designed because they affect the plant dynamics in terms
of magnitude and phase delay. If possible, they are chosen or designed so that their
bandwidths should be higher than that of the plant. If they have dynamic chractersitcs
in an operating frequency range, the order of G(s) can be more than 2nd order though
the plant is second order mechanical sytem. If they have little dynamic chractersitcs,
the 2nd order mechanical system G(jw) composed of a mass, damper, and spring can
2
ωn
be represnted as G(jω) = ω2 −ω2 +j(2ζ ω
)
and it is shown in a Bode plot as in Fig.6.2.
n ωn p
As addressed in Section 5.3, Chapter 2, It has a peak magnitude at ωr = ωn 1 − 2ζ 2 ,
Mr = |G (j )|max = √1 2 .
2ζ 1−ζ

109
Figure 6.2: The feedforward transfer function G(s) with little dynamics of the amplifier
and sensor when the plant is the 2nd order mechanical system.

Q: How should G(s) be desinged for a good tracking performance in a feed-


back system? Which elements among the controller components shown in
Fig. 6.1 contribute for this purpose?

Another representation method of G(jw) is shown in a polar plot whrere its magni-
tude and phase are drawn in the real and imaginary axes. One example of the 2nd order
C(iω) 2
ωn
mechanical system represented by G(jω) = R(jω) = ω2 −ω2 +j(2ζ ω
)
is shown in Fig.6.3.
n ωn

110
Figure 6.3: Polar plot of G(s)

K
Q: Draw a |G (jω)| = ms2 +bs
in a polar plot.

3.1 Voltage amplified loop

Figure 6.4: Voltage amplified loop

Depending on the types of amplifiers, we have different the feedforward loops. If a


voltage amplifier is used for driving an actuator such as a electrical motor, the feedforwad
loop is configured as shown in Fig.6.4. The electrical equation associated with the
electromechanical system such as a motor is obtained in
di
V =L + Ri + eb (6.1)
dt
Here, V and i are the voltage and current applied to the motor. L and R are the
inductor and resistor of the motor. eb is the back emf(electromotive force) induced when
the velocity of the motor v is generated in the magnetic field. The current i of the

111
motor is related to the force F generated by the motor. This phenomenon is called an
electromechanical principle. In mathematical expression, we have the following relations
F = Kt i (6.2)
v = Ke eb (6.3)
where Kt and Ke are defined as the force constant and electromotive force constant re-
spectively. Since there is internal velocity feedback loop, the velocity v will be controlled
to be constant by the voltage applied V . For example, when F happens to be increased
v will be increased. Then, eb proportional to v is also increased. However, i will be
decreased because of a small amount of V − eb , which finally decreases F . Hence, v will
be regulated again to be the same as before it is increased.

The bond graph corresponding to Fig. 6.4 is represented in Fig. 6.5. It shows v is
regulated again to be the same as before when F happens to be increased due to the
crossing relation of Eq. (6.2) and Eq. (6.3).

Figure 6.5: The bond graph corresponding to Fig.6.4


.

When a robot is operated using a voltage controlled loop, the block diagram showin
in Fig. 6.4 has the internal velocity feedback loop resulted from the relations of Eqns.
(6.1), (6.2), and (6.3). Hence, the robot maintains constant velocity that is proportional
to the input voltage V . However, when an end effector of the robot happens to be
stuck to an obstacle that prevents it from moving with the constant velocity, the motor
will draw current until the robot moves with velocity proportional to the input voltage.
Then, the motor will be eventually burnt out because of the increased current supplied
from a power supply. Therefore, it can be said that this voltage amplifed loop should
not be used when it happens to contact to other objects.

112
3.2 Current amplified loop

Figure 6.6: CurrentControlledOpenloop

When a robot is operated using a current amplified loop as in Fig.6.6, the block diagram
has an no feedback loop. Therefore, there must be a steady state error in position unless
the control components are correctly modeled. It provides constant force to a robot.
When an end effector of the robot happens to be stuck to an obstacle, the motor will
flow current until it delivers a designated force. It will not consume the current larger
than that required for the force. It can be said that this current amplified loop should
be used when it happens to contact to other objects. However, it can cause a steady
state error if there are modeling uncertainties Therefore, a position feedback loop can
be consdiered to reduce a steady state error.

4 Closed loop system analysis


4.1 Conventional controllers
To enhance the system performance such as transcient response characteristic and steady
state error, a controller can be added. Its performaces can be analyzed in a time domain.
Let’s talk about the transfer function G(s) whose input and ouput are the force and
position for the 2nd order mechanical system composed of a mass m and a damper
chracterized by c. It has a drift problem due to the integrator which causes collision to
an end. Then, G(s) is
K
(6.4)
s(ms + b)

Q: Where can we find in real applcation of the transfer fucntion expressed


G(s) in Fig.??.

113
A feedback system using a P controller C(s) = Kp can be constructed to solve this
problem as shown in Fig.6.7.

Figure 6.7: A proportional controller

The transfer fuction of the system with the P controller Gcl (s) is represented as

C(s)G(s) Kp K
Gcl (s) = = (6.5)
1 + C(s)G(s) ms2 + bs + Kp K

Q: What is the difference in G(s) and Gcl (s)? Investigate what advantages of
the feedback system can take?

A feedback system using a PD controller C(s) = Kp (1+Td s) is represented in Fig.6.8


to enhance the system performances.

Figure 6.8: A PD controller

Q: How is the transfer function of the system with the PD controller Gcl (s)
is represented? Please figure out how the closed loop system with the con-
troller is changed in terms of damping ratio and damped natural frequency.

C(s)G(s) Kp K(1 + Td s)
Gcl (s) = = 2
(6.6)
1 + C(s)G(s) ms + (b + Kp K + Td )s + Kp K

As indicated in Eqns. (6.5) and (6.6), a high Kp and Td cantribute to make a controlled
system have a high natural frequency, namely a high bandwidth. However, a high Kp

114
sometimes causes the system unstability, hence, a proper gain should be selected be-
tween the bandwidth and unstabilty points of view.

Q: How is the PD controller implemented?

Vo R2 R4
= (6.7)
Vi Z1 R3

= R2 R
R1 R4 (1+R1 C1 s)(6.8)
3

= Kp (1 + Td s)(6.9)

From the characteristic equation in Eq.??, we can make analysis of the system per-
formances using the relations of ωn2 = KKP , and 2ζωn = (b+KKP +Td ) as introduced in
Section ??, Chapter 2 . A desired natural frequency and a damping ratio are determined
using the controller gains Kp and Td . The effects of ωn2 and 2ζ on the system response
are analyzed in the s plane(σ − jω) where the roots of the characteristic equation s1 and
s2 are indicated. The roots s1 and s2 are varied depedning on Kp and Td and they are
called root locus.

115
The advantage of the time domain closed loop analysis is that it is possible to make
analysis mathematically. The controller gains of C(s) can be correctly determined by
investigating its output response. However, It is really complicated and tedious job.
Moreover, the output response should be correspondingly investigated for cases of dif-
ferent inputs.

4.2 Physical meaning of P, PD, PI controllers in a time domain


analysis

Figure 6.9: ControllerWithPlantLoop

When x is smaller than r the error e = r − x is positive. The postitive controlled output
is generated becasue the controlled output u is proportional to e. So, it makes x to
become larger Then. e will be zero when x = r. However, the inertia force the plant
makes x larger than r, so the negative e generates negative controlled output, which
again makes x smaller than r. Hence, this kind of the positive and negative controlled
output repeats until x is equal to r. Hence, it has a large overshoot and it takes a time
to reach to a steady state value as shown in Fig.6.10(a). When PD controller is used
de
dt
is added to e as shown in Fig.6.10(b). The D controller action provides a negative
input(braking force) u to a plant earlier than that in the P controller. This negative u

116
helps the plant to have a better dynamic characteristic which brakes the plant earlier.
Therefore, the PD controller usually can make the plant to have a wide bandwidth with
a smaller overshoot. The reduced phase delay due the differentiator can also make the
plant more stable. However, it can cause the feedback system to increase noise.

A PI controller has an integral action on the error e. The I controller action provides
a positive input(braking force) u to a plant later than that in the P controller. This
positive u can make the plant unstable. The advantage of using a I controller can be
found in an application, where there is friction associated with the motion of the plant.
When the controlled output u is smaller than the friction, we have a persistent steady
state error. However, this error is magnified by the integration. Then, the magnified
error is applied to a plant because it will be eventually larger than the friction force.
Therefore, it makes the plant move, which contributes to reduce the steady state error.

Figure 6.10: The differnt shape of e for the P, PD, and PI controller

117
4.3 Closed loop representation with disturbances and noise

Figure 6.11: A ball moving with the friction force

There exist other inputs in the closed loop system in addition to the command in-
put. Let’s suppose there is a ball moving on the floor by the applying force F with the
frction force Ff as shown in Fig. 6.10. The corresponding dynamic equation is expressed

F = ms2 + Ft (6.10)
When a position feedback system is considered to move to a desired distance, the
friction force Ff is anthor input to the plant represented based on Eq. 6.10 as shown in
Fig 6.12.

Figure 6.12: A feedback system with the friction force input

x(s)
Q: What is the transfer function Ff (s)
? When there is an external vibration
from the floor where is it added?

118
When a force feedback system is considered with a controller to apply a desired force
Fd , a block diagram can be constructed similary as shown in Fig. 6.12 except the sensor.
Since the applied force is required to be measured, a force sensor is used instead. When
a rigid needle is approaching to a rigid object, for example, it moves and contacts with
the body until the contact force is the same as the desired force Fd as shown in Fig. 6.13
with the solid line. In order not to break the needle, the plant applying the force should
widthdraw as quickly as it experiences a force larger than the desired force. In other
words, the controlled system should have a high bandwidth to reduce the force settling
time. However, as explained in the above, a high bandwidth requires a high controller
gain, which leads to the system unsatbility. To solve the problem, a flexible needle is
more peferable since it can decrease the control unstability. Its time response is shown
with a dotted line for comparing with the time response for the rigid needle in Fig. 6.13.

Figure 6.13: Time responses of force when a rigid and flexible needles are used for contact

In addition to the friction force represented as a disturbance in Fig. 6.12, we have


other input types of disturbances. When a force feedback system needs to be constructed
for regulating or controlling the contact force between an end effector and surface, the
variable gap distance is considered as an input disturbance to the plant as shown in Fig
6.14. Referring to the above examples, we can see the disturbaces have various types of
not only disturbing variable but also information to be known

Figure 6.14: A feedback system with the variable gap distance input

119
5 Open loop system analysis
5.1 Controller design for mechanical systems
The problem of this closed loop analysis is that it is impossible to make analysis without
knowing C(s). In other words, it is impossible to know the output response unless C(s)
is not designed. If we can design C(s) independently of G(s). It is easy to design C(s)
since we know how G(s) needs to be compensated for the performance improvement in
term of stability, speed, and steady state error. When the feedforward transfter fucntion
is C(s)G(s) as showin in Fig. 6.9, the transfer fucntion of the closed loop system, Gcl (s)

C(s)G(s) G0 (s)
Gcl (s) = = (6.11)
1 + C(s)G(s) 1 + G0 (s)
The closed loop output x(t) is unstable when G0 (s) = −1. This condition can be
separated as |G0 (jω) |=1 and 6 G0 (jω)=-180◦ . Hence, it can be said that the closed
system is unstable when |G0 (jω) | = 1 and 6 G0 (jω) = −180◦ . Here, we define a gain
margin as reciprocal of the magnitude |G (jωpc ) | at the frequency where 6 G (jωpc ) =
−180◦ . A phase margin is the amount of additional phase lag at the gain crossover
frequency ωgc required to bring the system to the verge of instability. We have positive
gain and positive phase margins represented at a Bode plot and polar plot in below.

Figure 6.15: Closed System Stability

When G0 (s) = C(s)G(s) have a positive gain margin and posiive phase margin, the
closed loop system is stable. If a P gain in the controller is increased to have a better
system performance, G0 (s) may have a negative gain margin or negative phase margin,
which makes the closed loop system unstable. Hence, the controller gains should be
properly selected. We also need to take caution in increasing controller gains because a

120
high gain may cause a saturation problem in the amplifier.

Additionally, the closed loop system performances can be determined by the gain
margin and phase margin of G0 (s) as

ωbw = ωgc whenP M ≤ 90◦ (6.12)


ωbw = 2ωgc whenP M ≤ 45◦ (6.13)
PM
ζ = whenP M ≤ 70◦ (6.14)
100
1 1
Mr = |G (j ω)|max = p = (6.15)
2ζ 1 − ζ 2 2 sin( PM
2
)

Q: The second order mechanical system composed of mass, spring and damper
is always stable. Why is that? When does an open loop transfer function
2
G(s)= s2 +2nns+ω2 have a negative phase margin? Then, is the open loop system
n
unstable?

Q: Most of a real second order mechanical system have a phase delay larger
than 180◦ at a high frequency. What make it have this result?

Q: Model a mechanical system having a negative phase margin. Prove it is


unstable at a closed loop system at a simulation work

We use a controller C(s) to increase gain margin and phase margin of the plant G(s)
and to secure the system stability and performances. A PI controller is modified to
cut off an integrating role and make a DC value in a low frequency, and the modified
controller is usually called a lag compensator. A PD controller is modified to cut off an
differentiating role and make a DC value in a lhigh frequency, and the modified controller
is called a lead compensator. It can reduce the high frequency noise that is generated
by the differentiation of the PD controller. The magnitude ratio of the PID controller,
lag compensator, and lead compensator are shown in Fig. 6.16.

121
Figure 6.16: Magnitude ratio of a PID controller

The phase delay of the PID controller is shown in Fig. 6.17. By applying the phase
lead and lag characteristics of the controllers C(s), the phase of the plnat G(s) is properly
adjusted in C(s)G(s).

Figure 6.17: The magnitude and phae delay of the PID Controller

5.1.1 Example: Desgin of a lead compensator

Consider the system shown in Figure 6.18. When you draw a Bode plot, we can see it
has a sufficent phase and gain margins, thus a controller gain has a room for increase

122
though the phase margin will be decreased as a result of increasing the gain. Hence,
it is imporant to design a controller that gives sufficent phase margin as well as a high
gain to decrease the steday state error. Let’s suppose that the controller gain is first set
to 10 to provide a small steady state error. Then, from 10G(jw) = 40(jw(jw + 2)) plot
shown in Fig. 6.19, the phase and gain margins of the system are found to be 17◦ and
+∞ db, respectively. A phase margin of 18◦ implies that the close loop system is quite
oscillatory. +∞ db gain margin implies that a controller has still a reasonably high gain.
Thus, the specification calls for a phase margin of 50◦ at least. We thus need to find a
controller to give the addtional phase lead 32◦ . To achieve this goal, a lead compensator
is chosen and designed.

Figure 6.18: Control system G(s) = 4/(s(s + 2)

Figure 6.19: The Bode diagram for 10G(jw) = 40/(jw(jw + 2)).

There must be some procedures of try-and-error method to find out cut off frequencies

123
with the consideration of the amount of the lead angle. Thus, the transfer function of
the lead compensator C(s) is designed with the determined gain and the additonal phase
as
0.2s + 1
C(s) = 10 (6.16)
0.05s + 1
Figure 6.20 shows C(s)G(s) provides the phase margin of 50◦ and high gain margin as
the result.

Figure 6.20: The Bode diagram for C(jw)G(jw).

To check the design rule of thumb suggested in Eqns. (6.13) and (6.14), the frequency
response of the closed loop transfer function Gcl (s) is obtained in Fig. 6.21. It is shown
that the closed loop bandwith is about 10 rad/sec, which is almost the same as wgc .

124
Figure 6.21: The VCM nanoscanner (a) schematic diagram, (b) the real configuration.

5.1.2 Control of a nano scanner used in AFM

Figure 6.22 shows the configuration of a VCM(voice coil motor) nano scanner used in
the AFM(atomic micro scope). It is composed of a mass, a leaf spring, a damper that
are typical elements of a mechanica system. The applying force is obtained from the
electromechncal principle obtained in the magnetic system composed of the permenent
magnet and electromagnet.

Figure 6.22: The VCM nanoscanner (a) schematic diagram, (b) the real configuration.

The frequency reponse of the nano scanner is shown in Fig. 6.23. Its transfer fucntion

125
G(s) is obtained using a curve fitting technique as

5.67109
G(s) = 4 + 51.43 + 3.571052 + 1.239106 + 1.885109
(6.17)

Figure 6.23: The frequency reponse of the nano scanner

As shown in Fig. 6.23, the phase margin is very small. A PD controller Kp (1 + T ds)
is used so that the phase margin and gain margin are properly adjusted. For exmaple,
if the controller gains Kp and Td are selsected as 0.2, 0.01, respectively, C(s)G(s) are
shown in Fig. 6.24. C(s)G(s) has the phase margin 47.7◦ and gain margin 3.3 db.

126
Figure 6.24: C(s)G(s) with a PD controller Kp (1 + T ds), Kp = 0.2 and Td = 0.01

When a PI controller Kp (1 + 1/Ti s) is used for Kp = 0.02 and Td = 0.05, C(s)G(s)


are shown in Fig. 6.25.

Q: How big bandwith of the closed loop system is expected for C(s)G(s) shown
in 6.24 from the design rule of thumb? How big bandwith of the closed loop
system is expected for C(s)G(s) shown in 6.25 ?

127
Figure 6.25: C(s)G(s) with a PI controller Kp (1 + 1/Ti s), Kp = 0.02 and Td = 0.05

Q: Investigate how a phase delay is improved and the system bandwidth


increases when a PD controller is used for a system composed of mass and
damper. Choose the parameters of G(s) on your own. Change Kp and Td to
provide a reasonable phase delay at a frequency of interest.

128
6 Controller design for electronic systems
Figure 6.26(a) shows an electronic feedback system called an operational amplifier. Its
block diagram is represented as shown in Fig. 6.26(b). A is a feedforward loop gain,
which is very large. β is a feedbak loop gain determined by resistors in the feedback
loop. There are various types of Op amplifiers in the market. They have different A and
the information is introduced in the data book.

Figure 6.26: (a) A feedback loop in an operational amplifier and (b) its block diagram.

One specific exmaple of an op amplifier is shown in Fig. 6.27. Since it is an open


loop transfer fucntion Aβ having a high gain, the op amplifier can be unstable when β
is not properly selected due to the negative phase margin.

129
Figure 6.27: An open loop transfer fucntion Aβ of a specific op amplfiers

Q: Find out how β is represented for the op amplfier circuit shown in Fig.
6.28.

Figure 6.28: One example of an op amplfier circuit

6.1 Design of control components for good command tracking


As seen in the above figure 6.29 on the next page, there are components in the closed loop
G0 (s)
system, that are Kc , Ka , Kt , G(s), and Ks . The overall transfer function YR(s)
(s)
is 1+G 0
(s)
.
We should have a condition of Y (s) ∼ = R (s) for good command tracking. It would be
realized if G (s) has a large magnitude. Ideally, if G0 (s) is infinity, Y (s) = R (s) , Then,
0

130
Figure 6.29: Components of the colsed loop system

does the increase of Kc , Ka , and Ks help the plant improve command tracking?

Q: Suppose that Kc = 1, Ka = 1(A/V ), and Ks = 1(V /m) for a system hav-


0
ing Kt = 1 N/A, G (s) = 1(m/N ). Then, G (s) = 1. Let’s compare with the case
V
, for systems having Kt = 0.1( N

Kc = 1, Ka = 10(A/V ), and Ks = 1 m A
), G (s) =
N
1(m/N ) and Kt = 1, ( A ), G (s) = 0.1(m/N ). We have the same overall transfer
function YR(s)
(s)
mathematically at the three cases. Which are the better cases
for good command tracking?

In simulation, threre is no differnecs in the results of the three cases in terms of


system performacnes such as stability and bandwidth because they look like haivng the
0
same G (s) . Here is the trap for doing the control simulation work. The high gains of Kc
and Ka (electrical gain) can cause instability and saturation problem when these kinds
of nolinear phenomena are implemented in the simulation works. We can not recognize
the effects can also make the coil in the actuator burn out. It is really important to
understand the voltage and current play key roles of providing good performance of the
plant because the back emf voltage and current are proportional to the velocity and force
in the plant. Hence, a maximum voltage and maximum current should be considered for
good command following even in the simulation. It is important to design the actuator
and plant to have high gains of Ka and G(s) unless they cause unstability.

131
Figure 6.30: A trapezoidal motion profile

Figure 6.30 shows a trapezoidal motion profile of a plant in a time domain. Since
the current in an electromechnical actuator is proportional to the force generated, it
determines the maximum acceleration that is the slope of the motion profile. Since the
voltage in an electromechnical actuator is proportional to the back emf generated, it
determines the maximum velocity. Suppose that the output v(t) for the ramp input r(t)
is obtained with little time delay from velocity control as shown in Fig. 6.30(a). The
distance travelled with this profile is assumed to be S. How does v(t) respond for the
inputs shown in Fig. 6.31(b) and Fig. 6.31(c)?

Figure 6.31: Reponses for the different trapezoidal profiles

The output respnoses are shown with dotted lines. The frequency of the ramp input
in Fig. 6.31(b) is higher than the input shown in Fig. 6.31(a). Therefore, the output
response is much delayed. The frequency of the ramp input in Fig. 6.31(c) is lower than
the input shown in Fig. 6.31(a). Hence, the output response follows well the input.
However, the output response is saturated becasue the ramp input is larger than the
maximum velocity. So, when a longer distance is required to travell, a longer span of

132
time is required.

Q: What type of an input should apply when a shorter distance than S is


required to move.

7 Dynamic effect of sensing system on control


The sensor is located after the plant to sense the physical properties of a plant as shown
figure 6.32. Here, Ks is a sensor gain, which is assumed to be constant. However, when
it is used for feedback system we need to take more attention on its dynamics because
it plays a role of increasing a phase delay.

Figure 6.32: Feedback loop with sensor

Figure 6.33 shows a frequency response of a ordinary sensor. Though the sensor gain
can be assumed to be constant at a low speed it is not constant any longer at a high
speed. The magnitude is much reduced at a high speed. Furthermore, it usually has
a phase lag as expected in its frequency response as shown in Fig. 6.33. The dynamic
effect of a sensor affects the system stability and performances. For example, a large
phase lag of a sensor can cause the system instability. In conclusion, it is necessary to
design a sensor whose bandwidth is much wider than system operating bandwidth. It
is recommended that the sensor bandwidth should be 10 times larger than the system
operating bandwidth.

Q: Show how the sensor bandwidth affects the closed loop system band-
width and stability using a simulation job. You can use any plant reasonably
modeled on your own. Keep in your mind that you should carefully choose
the amplifier gain, force constant, and sensor gain on the basis of realistic
application.

133
Figure 6.33: Frequency response of a normal sensor

8 Sampling effect on control


It is usually good to have a fast sampling in digital control because signal loss does not
happen. Suppose that 0.5 Hz sinusoidal signal needs to be detected at a sensor two
second is required to wait to collect the data for recognizing the signal at least.

If a digital sampling frequency is 1 Hz for example, the sinusoidal signal is recognized


as a DC value. If a sampling frequency is 2 Hz it is recognized as a ramp signal at best.
Therefore, it is recommended to sample the sensor at a rate faster than the signal
frequency or 5 ∼ 10 times faster in order not to distort the signal. However, it is not
always true that a high sampling frequency is always good for a system because a high
frequency signal is also detected at the sensor and it excites the plant, which causes
vibration for example due to the resonance effect. The high sampling frequency makes
it diffcult to realize a real time control becasue of a large amount of data, of course.

134
9 Loop shaping controller
If we include real disturbances a, w, v to the variables u, y we define eb,û. Here, ê = r − y,
u = û + a, y = z + v. r is a command, a is an input disturbance such as friction, w is
an output disturbance such as vibration, v is sensor noise. Then, a closed loop system
with real disturbances is represented as

The output we want to control is z.


z= Gu + w
y= z + v
One of control objectives apart from stability is to make z follow r (t) .
Therefore, consider the true error signal e defined to be
e= r − z
ê = r − y = r − z − v = e − v
The signal ê is used as the driving signal for the controller K(s)
We shall now analyze the loop equations in more detail.
u = û + a = K ê + a = K (r − z − v) + a
Therefore,

z = Gu + w = GK (r − z − v) + Ga + w
z = (I + GK)−1 GK (r − v) + (I + GK)−1 Ga + (I + GK)−1 w

The true error signal e is given by


e= r − z
={I − (I + GK}−1 GK} r + (I + GK)−1 {GKv − Ga − w}
=(I + GK)−1 r + GK(I + GK)−1 v − (I + GK)−1 Ga − (I + GK)−1 w
We call GK an open loop gain transfer function. Define (I + GK)−1 as a sensitivity
transfer function, S and GK(I + GK)−1 as a complementary transfer function, T . Then
we have

135
S+T =1
Note that A(I + A)−1 = (I + A)−1 A, (I + GK)−1 K.
Then, the true error signal e is represented as
e = S (r − w − Ga) + T v

10 Control Applications
10.1 Principle of AFM
Atomic force microscope (AFM) is a very useful instrument for measuring the topol-
ogy image of nano-scale samples in various industry and research areas such as MEMS
and Bio. In addition, AFM is used to measure the physical, electrical, and magnetic
properties, as well as the topology of images.
Figure 1 shows the configuration of the AFM system (Nano Xert-, EM4SYS). The
main components of the AFM are the laser diode, optical sensor (QPD), XY and Z
nano scanners, and cantilever. As the cantilever tip approaches the surface using the
PZT-driven Z nano scanner, it is deflected by an atomic force between the cantilever tip
and the sample surface. This deflection can be conventionally measured using an optical
sensor. When the XY nano scanner moves in the x and y directions, the atomic force of
the cantilever deflection is controlled in the z direction to maintain a constant reference
force Fd. Hence, displacement between the cantilever tip and the sample surface is
controlled to be constant by the z-axis force control. The presented configuration is our
experimental environment and our proposing method can be applied also in case of other
AFM configurations.

136
Figure 6.34: The configuraiton of an AFM system

The sample height is measured by measuring the deflection angle of the cantilever
using the optical lever principle as shown figure 2. The deflection angle of the can-
tilever is varied by the sample height as indicated Eq. 1. h/2l=sin(/2) (1) Where, h
is the height variation of the sample, l is the length of cantilever and is the deflection
angle of the cantilever. Eq. 1 can be simplified as indicated Eq. 2 under assumption
that the is closed to zero. h/l,when 0 (2) Therefore, the sample height variation h
can be obtained by measuring the cantilever deflection . The cantilever deflection an-
gle is magnified by the optical path length l0 asindicatedEq.3andEq.4.d = l0 tan2(3)d =
l0 tan(2hl)(4)W here, distheref lectedlaserlocationontheQP D, l0 istheopticalpathlength.T heref ore, thea
FIG. 2. Principle of the optical lever

10.2 Gap control in AFM


Figure 3 shows a block diagram of the force control loop used in the AFM. Height
variation in the sample can be considered as a disturbance and represented as Zd. The
position of the PZT-driven Z nano scanner is represented as Z. The optical sensor Ks
measures the change in the deflection angle of the cantilever, which is proportional
to the relative displacement Z that is obtained by subtracting Zd from Z. Here, Z is
proportional to the controller output u under the assumption that no dynamics exist in
the amplifier Ka and the Z nano scanner P(s). Therefore, Zd is subsequently obtained
by u if Z is controlled to be constant as the reference force Fd.
FIG. 3. AFM-control block diagram for the z-axis force control
Figure 4 (a) and (b) shows the SiC graphene topology image and its profile line,
respectively. It has the sub-nm height variation.

137
(a)
(b)

Figure 6.35: (a) The SiC graphene topology image and (b) its profile line

FIG. 4. SiC graphene (a) topology images (b) profile line

10.3 Sensors and actuators in AFM

138
Chapter 7

Electric circuit design and analysis

1 Introduction
First, I want to start this section asking a question whether the electrical variables such
as voltage and current can be applied to electrical components such as a resistor or
capacitor at the same time as we want. Let’s put aside the answer for the question for
a time being and we keep on going with a more easy work before we find the answer.

Figure 7.1: An electrical curcuit composed of a power supply, a resistor R and a capacitor
C.

Fig. 7.1 shows a simple electrical curcuit composed of a power supply, a resistor R
and a capacitor C. A circuit theory can be applied to analyse the circuit by applying
the well known Kirchhoff’s law[??].
Z
1
ein = RiR + iC dt (7.1)
C

139
Here, ein , iR , and iC are the applied voltage from the power supply, the current flowing
through the resistor, and the current flowing through the capacitor.

The second question is how Eq.(7.1) is obtained and how it is solved? The third
question is where this circuit is used and for what? The last question is that there isn’t
any way to get a universal physical concept and meaning from the electrical system for
other engineering students than electrical engineering students?

2 Power and signal


It was stated in Ch. 2 that there are power variables and energy variables, namely,
effort, flow, momentum, and displacement in electrical systems and mechanical systems.
Figure 7.2 shows one of examples showing how the applied electrical power ei is delivered
to the mechanical plants in an industrial robot, converted with T ω, f v, and fs vs . When
the electrical motor generates the current i from the applied voltage, the generated
electromechancal torque T is applied to the robot arm. Then, it rotates the robot arm
at the angular velocity ω. The translated velocity v from ω lifts the mass. The power
flow is simply represented using the block diagrams and half arrow in Fig. 7.3.

140
Figure 7.2: Applied electrical power delivered to the mechanical plants in an industrial
robot

Figure 7.3: The electrical power flow represented using the block diagrams in an indus-
trial robot

In the above statement, though v is only mentioned, we know the lifting force f is
also generated at the same time with v in order to produce the power. If the input
power is not sufficently applied, the weight will not be lifted wth having a desired
velocity. Therefore, the applied input should be well designed to carry out the desired
performance of the robot. However, it is commom to encounter a case in reality that a
power loss occurs while the applied power is tansmited through the plant components
due to power dissipated loads. Hence, It is sometimes necessary to measure them to
monitor if the power is well delivered to other plants or loads as intended for the plant
desgin purpose. Various commercial sensors are facilitated to measure them using an

141
electronic signal such as voltage or current. The reason of being called a signal is it has
no power, in other words, it does not have both but only one differently from a power
in which an effort and a flow go along together. The signal corresponding to the force
or velocity is denoted using a full arrow in Fig. 7.3.

3 Electric circuit modeling


3.1 Constitutive relations
It was stated in Ch. 2 that there were only 3 components in electrical systems, they
were an inductor L, a resistance R, and a capacitor C. There are four variables, volt-
age e, current i, flux linkage λ, and charge q that are equivalent to force f , velocity v,
momentum p, and displacement x in the mechanical system. e and i are called power
variables because product e and i is a power. The power variables and energy varaiables
of varioust physical systems are listed in Table7.1.

Table 7.1: Generalized variables in physical systems

effort(e) flow(f ) momentum(p) displacement(q)


Mechanical system f v p x
Electrical system e i λ q
Hydraulic system P Q Γ V
Magnetic system M φ̇ φ

We can recognize there are similarities among variables in those physical systems.
When damping force and spring force are indicated as fd and fs respectively, the four
variables are linearly represented as

p = mv, (7.2)
fs = kx, (7.3)
fd = cv, (7.4)

In other words, The mass m is a linear paremeter that relates monentum to veloc-
ity(flow). The spring coefficent k is a linear paremeter that relates force(effort) to
displacement. The damping coefficent c is a linear paremeter that relates force(effort)

142
to velocity(flow). Equations(7.3), (7.4), and (7.4) are called linear constitutive relations.
Of courese, those parameters and variables can be nonlinear at a specific system such
as a fludic system.

Similary, the definitions made in a mechanical system can be also applicable to the
electrical systems. In the electrical system we have the following constitutive relations.

λ = LiL (7.5)
1
ec = q (7.6)
C
eR = RiR (7.7)
An inductor L can be defined as a parameter that relates the monentum λ to current
through the inductor, iL . A capacitor C can be defined as a parameter that relates
voltage ec to charge q. A reisistor R can be defined as a parameter that relates voltage
eR to current trough the resistor R, iR . In other words, the definitions of L, R , and C
can be respectively made using the four variables voltage, current, momentum, charge
using the constitutive relations. We also have relations of λ and i from definitions the
momentum and flow that are obtained by differentiating the effort e and displacement
q respectively,
dλ dq
eL = , ic = (7.8)
dt dt
Hence, Eqns.(7.5), and (7.6) can be rewritten as

diL
eL = L (7.9)
dt
Z
1
eC = iC dt (7.10)
C
The above constitutive relations are graphically drawn in Fig.7.4.

143
Figure 7.4: The constitutive relations that are the same as the result obtained by ap-
plying the Kirchhoff’s law.

Q: What is the answer for the first question? Please validate your answer.

3.2 Bond graph model of electric circuits


For representing the constitutive relations in a more general form which shows similari-
ties with other physical systems, a bond graph is introduced as showin in Fig. 7.5. The
graphical notaions of the capacitor, resistor, and inductor are denoted using the port
C, R and L instead. The power varables are denoted at the upper and lower parts of
the bond. The half arrow notation is used to express the power flow direction. The bar
notation is used to include the causality relations among four varables. For example, in
the C port, the current through the capactor ic is applied, then the capacitor voltage
ec is generated in return. In the R port, either current or voltage can be applied, then
voltage or current is respectively generated in return. In the L port, the voltage eL is
applied, then current iL is generated in return.

144
Figure 7.5: A bond graph representation of electrical components

For a reference, the electrical energy E is defined by integrating the electrical power
P with respect to time. Hence, E is represented using the constitutive relations as
Z Z Z Z
dq 1 2
E = P dt = eidt = e dt = edq = q (7.11)
dt 2C
λ2
Z Z Z Z
dλ 1
E = P dt = eidt = i dt = idλ = = Li2 (7.12)
dt 2L 2
Since the electrical power P is a product of force e and velocity i, F and v are called
power varaibles. Similarly, since the mometum λ and displacement q are appeared in the
energy relations as indicated in Eqn.(7.12) and (7.12), λ and q are called energy variables.

Recalling some rules of the bond graph representation, the electrical circuit of Fig.7.1
is represented using a bond graph as shown in Fig. 7.6.

145
Figure 7.6: The bond graph representation of the electronic circuit

If the voltage at the points 1,2,3, and 4 are denoted by e1 , e2 , e3 , and e4 , e1 = ein since
there is no voltage drop in the conductor. Likewise, e2 = e3 , e4 = 0. If the current at
the points 1,2,3, and 4 are denoted by i1 , i2 , i3 , and i4 , we know that iin = i1 = iR =
i2 = i3 = iC = i4 . The 1 juction is enforced to show this condition. The voltages around
the 1 juction has the relation of ein = eR + e2 from the Kirchhoff’s voltage law, and it is
represented by the half arrow directions of the in-going bond and out-going bonds. In
other words, the voltage ein of the in-going bond is the same as the summatiion of the
two voltages eR and e2 , the out-going bonds. We also have e2 = e3 = eC because e4 = 0.
Thus, we have following two conditions,
iin = iR = iC = i (7.13)
Z
1
ein = eR + eC = R1 i + idt (7.14)
C
From the above relations, the bond graph shown in Fig.7.6 is the equivalent to the bond
graph shown in Fig.7.7.

Figure 7.7: An equivalent bond graph showin in Fig.7.6

146
Figure 7.8 shows another circuit having a parrell connection with the previous circuit.
Figure 7.9 shows its bond graph representation. The bond graph shows how the power
or energy associated with the parameters L, R, and C are delivered in the system. In
the electrical circuit, we know that i2 = iC + i3 from the Kirchhoff’s current law, that is,
the current summation at the juction equal to zero. The 0 juction is enforced to show
this condition. The voltages around the 0 juction are the same each other. In summary,
it can be said that the Kirchhoff’s voltage law and current law are respectively imple-
mented using the 1 junction and 0 junction.

Figure 7.8: An example of an electronic circuit

Figure 7.9: The bond graph of the electronic circuit shown in Fig. 7.8

First, the voltage input ein is the applied to the electric circuit and it is represneted
as an effort source in the bond. Second, the net voltage (ein − e2 ) acting on the resistor
R1 is represetned in the bond using 1 port. Then, the current iR1 is obtained as a result

147
of the net voltage eR1 applying to the resistor R1 . Third, the voltage e2 with current
i2 is delivered to the next component C port. Fourth, i2 is divided into two different
paths providing iC and i3 . iC is applied to the C port and generates the voltage eC as
a result that is obtained using the constitutive relation qCC . Lastly, e3 is applied to the
next elements R2 and L. The voltage e3 is used to supply e4 acting on the L element.
Then, iL is obtained as a result using the constitutive relations associated with L. Then,
the current iL that is the same as iR2 is applied to the R2 element. Finally, the current
iR2 is determiend as a result using the constitutive relation associated with the resitor
R2 . One thing to note is that the inductor, resistor, and capacitor are acting as loads
to the applied voltage or power. Hence, voltage or power will be eventually decreased
as they are encountered with these kinds of loads.

In conclusion, from the 1 junction and 0 junction charactersitics, we have following


conditions.

iin = i1 = iR1 + i2 , (7.15)


ein = eR1 + e2 (7.16)
i3 = iR2 = iL (7.17)
e3 = eR2 + e4 (7.18)

From the constitutive relations, we have


eR1
iR1 = , (7.19)
R1 Z
1 1
eC = qc = iC dt (7.20)
C C
eR2 = iR2 R2 (7.21)
Z
λ 1
iL = = eL dt (7.22)
L L
Solving the above differential equations mathmatically in a time domain, we can figure
out how the variable of interest in the electrical circuit behaves when the input applies.
It does not seem to be easy to calculate them though.

Q: What is the voltages at the point 1, 2, 3, and 4 in Fig.7.8?

148
4 Time domain analysis
The electrical equation for the circuit shown Fig. 7.6 is represented again as
dein di 1
=R + i (7.23)
dt dt C
Equation(7.23) is the first order differential equation that is solved at two seperate cond-
tions of the trancient state condtion and steady state condition. The transcient response
is obtained assuming dedtin = 0. The steady state response is solved assuming that dedtin
exists. Then, initial conditions are applied to the generalized solution combining the
transcient response and the steady state response to obtain the final solution.

di
At the transcient condtion that will last only for a short time, RC dt = −i from
Eq.(7.23) assuming ein =0. Applying the varaible seperation method, we obtain

di 1
=− dt (7.24)
i RC
Then, by integrating both sides,
1
ln i = − t + k, (7.25)
RC
Here, k is an integral constant. Then, we obtain the transcient response as

1
i(t) = e(k− RC t) (7.26)

At the steady state condition that will last for a long time, more complicated process
is usually taken to find out i. However, when ein is constant in this case, the steady
di
state response is simply obtained by assuming that dt = 0 from the definition of the
steady state condition, that is, i is eventually constant at t = ∞. Then, we have i = 0
since dedtin = 0. When the initial conditions i(0) = c are applied, the final solution of
Eqn.(7.23) is
1
i = ce− RC t (7.27)
The current i is shown in Fig.7.10.

149
Figure 7.10: Current in an RC circuit when the input voltage is constant

If ein = eo sin ωt, then


dein
= ωeo cos ωt (7.28)
dt
Through a mathmatical process[Advanced Engineering Mathematics], we obtain i(t) as

−t ωeo C
i(t) = ce RC + (cos(ωt) + ωRC sin(ωt)) (7.29)
1 + (ωRC)2
Figure 7.11 shows i(t) in the RC cirucit for the sinusoidal input.

Figure 7.11: Current in an RC circuit when the input voltage is sinusoidal

More complicated analysis will be considered at RLC circuits since they contain a
resistor, inductor, and capacitor, which is represented by 2nd order differential equation
as in the mechanical system.

The time domain analysis of the dynamic systems represented using the differential
equation is usually very complicated as expressed in the above process because the dif-
ferential equation needs to be solved with the inpult specified. The analysis will be more
complicated especially when the system is a higher order dynamic systems. It is almost
impossible to obtain the analytical solution in this case. These reasons lead to use the
frequency analysis of the dynamic systems. Let’s find out how it is differed and how it
is beneficial to the system engineer.

150
Instead of solving the differential equation, a Laplace transformation method[Modern
Control Engineering, Ogata, ch.1,section 3] is used to takes the adnvatage an operational
method of solving linear differential equations. Applying it to both sides of Eq. (7.14),
that is the integral expression of Eq. (7.23) we obtain with the assumption of i(0) = 0.
1
Ein (s) = R1 I(s) + I(s) (7.30)
Cs
where Ein (s) and I(s) are the Laplace transformed functions of ein (t) and i(t) respec-
tively. The second term of the right side of Eq.(7.30) is obtained by taking the procedure
of

Z t  Z ∞ Z t 
L i(t)dt = i(t)dt e−st dt (7.31)
0
0Z t 0  −st Z ∞
e ∞ e−st
= i(t)dt |0 − i(t) dt (7.32)
0 −s 0 −s
1 ∞
Z Z
1
= i(t)dt|t=0 + i(t)e−st dt (7.33)
s s 0
i−1 (0) I(s)
= + (7.34)
s s
where i−1 (0) = i(t)dt.
R

For the second order differential equation, refer to section 5. Though it is regarding
to the 2nd order system of a mechanical system, the anaysis can be also applied to the
2nd order electrical system if the parameters m, c and k are equivalently replaced by
L, R, and 1/C.
One of big advantages of using the Laplace transformation is that the dynamic equa-
tion in the time domain can be transformed into an algebraic equation in a s complex
plane. Using this favor, the relation of an output versus an input can be easily obtained
using a transfer function, EEin
C (s)
(s)
defined as the ratio of the Laplace transformation of the
output to that of the input under the assumption that all initial conditions are zero.
When the output is the voltage of the capacitor ec (t), Ec (s) is obtained as

1
EC (s) = (7.35)
Cs
EC (s)
Then, the transfer function of Ein (s)
is algebraically obtained

EC (s) 1/(Cs) 1
= = (7.36)
Ein (s) R + 1/(Cs) 1 + RCs

151
eC (t)
However, it is not possible to describe mathematically ein (t)
.

Equation(7.36) can be used to obtain the output EC (s) when the known input Ein (s)
is applied. Then, ec (t) is exactly obtained by using the inverse Laplace transformation
method. For example, when ein (t) = 1, EC (s) is obtained as Eq.(7.37) using the relation
R∞
of Ein = 1s (refer to Ein (s) = 0 1 · e−st dt),
1 1
EC (s) = · (7.37)
(1 + RCs) s
Thus, eC (t) is obtained using the look-up table of the inverse Laplace transformation as
1
eC (t) = 1 − e− RC t (7.38)

Another advantage of the Laplace transformation is that it allows the use of graphical
techniques for predicting the system performances without actually solving the differen-
tial equations, that is, without relying on the inverse Laplace transformation method.
This analysis is called a frequency response.

5 Frequency response analysis of an electrical sys-


tem
Let’s suppose that the input and output are denoted by x(t) and y(t) respectively. If
the sinusoidal input signal x(t) is applied to a plant or system, the output willl be also
sinusoidal signal of the same frequency but with possibly different magnitude and phase
angle. Here, we define a transfer function of a plant or system G(s) as the ratio of the
Y (s)
Laplace transformed output and the Laplace transformed input. Thus, G(s) = X(s) .

Figure 7.12: Stable linear time-invariant system

The sinusoidal input signal is mathematically expressed by

x(t) = X sin ωt (7.39)

152
Then, the output singal can be represented with the different magnitude and phase delay
by
y(t) = Y sin(ωt + φ) (7.40)
For example, Fig. 7.13 shows one of experimental result obtained at ω = 12rad/sec.
It shows the real input and output sinusoidal signals together having magnitude reduc-
tion and phase delay.

Figure 7.13: Input and ouput sinusodial signals having different amplitude and phase
delay

Y
The ratio of X and the phase delay φ are experimentally obtained and ploted with
solid dots in M and φ axis respectively with respect to ω ranging from a low frequency
Y
to a high frequency as shown Fig.7.14, which is called the frequency response. X and φ
are important indexes in analysing the dynamic characteristics or specifications of the
electrical circuit or mechanical plant because they provide the information how well the
output follows the input in terms of the magnitude and phase.

153
Figure 7.14: The frequency respnose obtained with experiment.

When the output behaves for the input as shown in Fig. 7.14, it has severe magnitude
reduction and phase delay at a frequncy higher than 10rad/sec. This result is not good
for target tracking applications. For example, when the robot arm moves the weight to
a target poistion because it does not reach to the target position. What is worse, it has
a delay. The most desired motion is achieved when Y = X and φ = 0. However, when
the input is external vibration from the ground and the output is the displacement of
the robot arm, the frequency response result would be favorable because the vibration
input is reduced in the output, which affects less the robot arm motion.

It was proven[Modern control engineering, Ogata] that the frequency response can
be also obtained directly from the transfer function G(s) with s replaced by jω. Here,
ω is the freqeuncy. Thus, we have the relaion of
Y (jω)
G(jω) = = M6 φ (7.41)
X(jω)
where M is the amplitude ratio of the output and input sinusoids. Here, the input
frqeuency ω is varied so that the frequency range of interest is covered. Equation (7.41)
indicates that the transfer function of the plant can be obtained using the frequency re-
sponse. In reality, the behavior of the simple RC circuit analytically obtainable by just
deriving the differential equation as expressed in Eq. (7.23). However, it is not easy to
figure out its behavior in the high freqeuncy because it has unexpected dynamics. There

154
are more severe problems in the mechanical systems such as wrong parameter estimation
in addition to the high frequency dynamics. In this case, the frequency response can be
applied to identity the model more correctly.

What will happen when the input is other than the sinusoidal? Can we figure out the
dynamics of a plant or system using the frequency response, for example, if the inputs
are constant or impulse inputs? Fortunetely, they can be represented by the summation
of sinusoidal inputs. If f (t) is a period function of period T , f (t) can be represented by
a trigometric series according to Fourier Series theorem.

f (t) = a0 + Σ∞
n=1 (an cos nt + bn sin nt) (7.42)
Z T /2
1
a0 = f (t)dt
T −T /2

When f (t) = 1, for example, Fig. 7.15 shows how the step function is composed of the
summnation of the sinusoidal fucntions.

Figure 7.15: A constant input represented by the summation of sinisoidal inputs having
varied frequencies.

5.1 Frequency response of the first order passive electrical cir-


cuit
Consider again the elctrical circuit shown in Fig. 7.6. The transfer function of the
elctrical circuit G(s) is represented as EEin
C (s)
(s)
with the consdieration of the input Ein (s)

155
and output EC (s). Then, the transfer funcion G(jω) is
1
G(jω) = (7.44)
1 + RCjω

The amplitude ratio M and phase delay φ of G(jω) are respectively obtained using a
log magnitude for the sake of scale representation as
1 p
M = 20 log |G(jω)| = 20 log | | = −20 log 1 + (RCω)2 db (7.45)
1 + jRCω
1
φ = 6 ( ) = − tan−1 RCω (7.46)
1 + jRCω
1
At a low frequency, such that ω  RC
, the log magnitude may be approximated by
p .
−20 log 1 + (RCω)2 = −20 log 1 = 0db (7.47)

Thus, the log-magnitude curve at low frequencies is the constant 0 db line. The phase
delay φ may be approximated by
1 .
6 ( ) = − tan−1 0 = 0 (7.48)
1 + jRCω
1
At a high frequencies, such that ω  RC
p . p
−20 log 1 + (RCω)2 = 20 log (RCω)2 = −20 log RCωdb (7.49)

The phase delay φ may be approximated by


1 .
6 ( ) = − tan−1 ∞ = −90◦ (7.50)
1 + jRCω
Finally, M and φ are are described in Fig.7.16 in a logarithm scale Bode plot with
respect to the applied frequency.

156
Figure 7.16: Frequency response of the first order electrical system

Form the frequency response, we know the magnitude of the output voltage eC (t)
is the almost the same as that of the input voltage ein and little phase delay at the
low frequency. However, the magnitude of the output voltage eC (t) is much reduced
and there is a phase delay of −90◦ at the high frequency in the electrical circuit. That
is why it is called a low pass filter. It can be used to reduce the magnitude of high
freqeuncy signal whose typical example is electrical noise. The high order low pass filter
can be constructed to reduce the magnitude more after the cutoff frequency though a
larger phase delay is generated. There must be a caution since a large overshoot can be
produced if the damping ratio is small.

1
Q: Draw the frequency response of the transfer fuction G(s) = 1+0.01s in
a logarithm scale. Sketch the output for the input x = sin 2π(10Hzt) and
x = sin 2π(100Hzt) respectively. Compare the results with the above results.

An experiment will be carried out by constructing a passive circuit composed of a


power supply, a resistor, and capacitor whose circuit model is shown in Fig. 7.6. A
function generator can be used instead of a power supply because it can supply reason-
able voltage and current and it is convenient to supply a sinuoidal input having variable
magnitude and frequency. Students are requested to prepare for a function generator,
electical components including a mother board, and an oscilloscope to measure input
and output signals. Figure 7.17 shows a configuration of a curcuit implementation using

157
a mother board.

Figure 7.17: A configuration of a RC curcuit implementation using a mother board

The frequeny response test should be done applying a sinusoidal input signal having a
constant magnitude. Measure the voltage across the capacitor for the variable frequency
input in terms of the magnitde ratio and phse delay. Check if the circuit analysys result
matches well to the selected components. Draw the magnitde ratio and phase delay
with respect to the frequency in a logarithm scale. Check when the magnitude ratio is
drastically reduced.

Report the experimental results with analysis two weeks later of the date the home-
works are assigned.

5.2 Frequency response of the second order passive electrical


circuit
Now, think about a RLC electrical circuit composed of a resistor, inductor, and capacitor
as shown in Fig.7.18 when the input is ein . If the voltage of the capacitor ec is an output,

158
the transfer function G(s) of the output to the input is obtained from a differential
equation of the RLC circuit and is represened by

EC (s) 1/Cs 1 1/LC


G(s) = = = =
Ein (s) Ls + R + 1/(Cs) LCs2 + RCs + 1 s2 + (R/L)s + 1/LC
(7.51)
The standard form of Eq. (7.51) is wriiten using ζ and ω as

EC (s) ωn2
G(s) = = 2 (7.52)
Ein (s) s + 2ζωn s + ωn2
q q
where ζ and ω are defined respectively by R2 CL and LC 1
.

Figure 7.18: A RLC circuit

Then, G(jω)) is represented using the general notations of the input and output as

Y (iω) ωn2
G(jω) = = 2 (7.53)
X(jω) ωn − ω 2 + j(2ζ ωωn )

The amplitude ratio M and phase delay φ of G(jω) are respectively obtained as

1 1
M = |G(jω)| = = (7.54)
1− ( ωωn )2 + j2ζ ωωn a + bj

1
φ=6 ( ) (7.55)
a + bj
M and φ are are described in Fig.7.19. They are described in a logarithm scale Bode
plot with respect to the applied frequency.

159
Figure 7.19: Frequency response of the 2nd order system

We can find that there are some characteristics in G(jw) shown in Fig. 7.16 and
Fig. 7.19 in which the order of s in denominator is greater than the order of s in the
numerator. The first thing is that no much change in amplitude is observed at a low
frequency while a large reduction in amplitude is observed at a high frequency. That
is why they are called low pass filters(LPF) which are interpretated as ’a low frequency
passing’ or ’following’ device. When the cutoff frequency is higher, the output signal can
follow the input well at a wide frequency range with little magnitide reduction. This is
a design goal of a high speed system. However, when the cutoff frequency is lower, the
output signal can follow the input at only lower frequency range. The output can not
follow well the input even though a high speed(frequency) input is applied.

160
If the input is noise, we need to think differently. SInce the noise should be prevented
from entering to a plant, it should be reduced as much as possible. Hence, a LPF needs
to be designed to have a lower cutoff freqeuncy differently from the case when siganl is
entered to a plant. Therefore, it is important to know where is the frequency region of
the signal and noise. Fortunetely, they have different regions, in other words, signals are
mostly living at a low frqeuency while noise dominates at a high frequency. the cutoff
frequency should be well selected so that signal survies and noise is elliminated. For
example, when there is a signal with noise shown in Fig. ??(a) in the time domain, a
LPF should be designed by checking the frquency spectrum shown in Fig. ??(b)

What is the differene between G(ω) of the RC electronic circuit and G(ω) of the
RLC electronic circuit? First of all, the attenuation rate after the cutoff frequency is
much greater compared with the RC circuit. That is why it is more effective in reducing
the high frequency noise to use the second order RLC circuit when electrical noise is
assumed to be almost in high frequency range. Secondly, the phase delay is as large as
180◦ . Hence, though it has the advantage of a greater redcution of the high frequency
noise, it has a disadvantage of a larger phase delay which might cause unstability a
feedback system. This effect will be studied in mofre detail in the next section.

Q: Draw the frequency dependent impedance defined as EIin (ω) for the elec-
trical circuit drawn in Fig.7.18 when the input is ein . When the output is
the voltage across the resistor eR what is the transfer fuction EEin
R
? Draw its
frequency response in a Bode plot.

1
Q: Draw the frequency response of the transfer fuction G(s) = 0.01s in a
logarithm scale. Sketch the output for the inputs x = sin 2π(10Hzt) and
x = sin 2π(100Hzt), respectively.

161
5.3 Specifications of steady state response characteristics
If G(jw) has a peak value at some freqeuncy, this frequency is called the resonant
frequency wr . A peak value of |G(jw)| will occur when

w2 2 w
g(w) = (1 − 2
) + (2ζ )2 (7.56)
wn wn

is a minimum. Thus, the resonant frequency wr is


p
ωr = ωn 1 − 2ζ 2 (7.57)

The magnitude of the resonant peak Mr can be found by substituting Eq. (7.57) into
Eq. (7.54). For 0 ≤ ζ ≤ 0.707,

Mr = |G(jw)|max = |G(jwr )| = √1 (7.58)


2ζ 1−ζ 2

Let’s investigate how the steady state response y(t) is obtained when a step input
is applied. We can find out y(t) diretly by using the inverse Laplace transformation of
Y (jω) represented in Eq.(7.53). We can also refer to the frequency response in obtaining
y(t). As mentioned in Section 5, we know that the step input is composed of summation
of sinusoidal inputs. Hence, when ω is less than ωn , Fig.7.19 indicates that y(t) follows
well the low frequency inputs. However, when ω is larger than ωn , y(t) does not respond
well to the high frequency inputs. When ω close to ωn applies, y(t) responds with an
increased magnitude especially when ζ is small. Therefore, we can expect y(t) bahaves
as shown in Fig. 7.20.

Figure 7.20: Frequency response of the 2nd order system

162
where Mp , tr , tp , and ts are the maximum overshoot, rising time, peak time, and settling
time.
Those performacne specifications can be easily found out mathematically from y(t)
using a relation, for example, Mp = y(tp ). We can also find out a relation between time
response specification and frequency response specification from the frequency of the
shape of the first part of time response curve y(t), that is
π
tp = (7.59)
wd
Furthermore, we have the following relations,
4
ts = , (2%steadystatecriterion) (7.60)
wn
3
ts = , (5%steadystatecriterion) (7.61)
wn
The system bandwidth ωBW is defined a frequency at which the magnitude of the
transfer function is 3db below its zero-frequency value. It is oftten called a cutoff fre-
quency. From its definition, the bandwidth is almost the same as the natural frequency,
that is, ωBW = ωn . Conceptually, ωBW is the maximum frequency at which the output
follows well the input signal in terms of the magnitude ratio and phase delay. Therefore,
the bandwidth of the plant is regarded as a speed performance index.

Q: We can find real examples of plant whose behavior can be understood us-
ing the frequency response analysis. What do you think about the bandwidth
of an automobile driven along the curved road with the speed of 100 (km/hr)?

163
Figure 7.21: An automobile driven along the curved road.

Q: What is the bandwidth of an airplane landing with the speed of 360


(km/hr) on the ground having a landing displacement 1km?

Q. Draw the time response of the 2nd order plant when ω = ωl (low fre-
quency), ω = ωn , and ω = ωh (high frequency).

Q: If we are asked for designing a low pass filter which can obtain the dis-
placement from the velocity signal sin(100Hzt) how big order low pass filter
is required? How big the cutoff frequency is determined? What will happen
if an ideal integrator( 1s ) is used instead of a low pass filter?

6 High pass filter


What will happen when the output ER is the voltage across the resistor in the electrical
circuit shown in Fig. 7.6? Then, the transfter function G(s) is obtained from the first
order differential equation associated with the RC circuit as

ER (s) R RCs
G(s) = = = (7.62)
Ein (s) R + 1/(Cs) 1 + RCs

Then, G(jω) is represented


ER (jω) jRCω
= (7.63)
Ein (jω) 1 + jRCω

164
We can draw G(jω) directly by plotting it with respect to ω ranging from a low
frequency and high frequency. Another way to draw G(jω) is that Eq. (7.63) is composed
1
of jRCω and 1+jRCω and we take the advantage that the cross product of the Laplace
transformed terms can be drawn in Bode plot by adding them analytically, which is one
1
of advantages of the Laplace transformation. After drawing jRC and 1+jRCω seperately
jRCω 1
and adding them, we obtain 1+jRCω finaly. Since we know that jRCω and 1+jRCω are
just a differentiator and first order low pass filter, they can be easily drawn with the
help of previous experiences as in Fig.7.22.
The high pass filter can be used to elliminate the low frequency signals such as the
electrical offset or DC disturbance that are mostly occured in a low frequency range
because it has an feature of a large reduction in the magnitude ratio before the cutoff
freqeuncy. Of course, the second order high pass filter can be also designed to elliminate
the low frequency signal with more reduction rate.

Figure 7.22: G(w) of a high pass filter

1 1 0.05s+1
Q: Draw the transfer functions, 0.02s , s, 0.0001s2 +0.05s+1 , and 0.0001s2 +0.05s+1
in a
Bode plot to show the magnitude ratio and phase delay.

Q: Sketch the output responses for the RC circuit and LRC circuit with a
damping ratio 0.3 when the step input and ramp input are applied. Suppose

165
that the input frequencies are lower than the cutoff frequencies of the circuits.

166
Chapter 8

Signal processing using operational


amplifiers

1 Passive system; non-isolated system


There was an electrical circuit composed of a resistor R and capacitor C with an voltage
input as shown in Fig.7.6 in Section 3.2, Chapter 7. Its governing equation was obtained
and described using the 1st order differential equation. If another circuit composed of a
resitor is serially connected to the previous circuit as shown in Fig.8.1, can each circuit
be described seperately like two isolated systems? Here, the two resistors are represented
with R1 and R2 for differentiation. In other words, is it possible to obtain the transfer
funtion EEino without considering R2 ? This question can be answered by derivng the
electrical cicuit.

Figure 8.1: An electronic circuit composed R1 , C, and R2

The dynamics of the electronic circuit composed of R1 and C are represented as

167
Z
1 L{v(t)} 1
1) v (t) = R1 i (t) + i1 (t)dt −−−−→ V (s) = R1 I (s) + I1 (s)
C Cs
L{i(t)}
2) i (t) = i1 (t) + i2 (t) −−−−→ I (s) = I1 (s) + I2 (s)
Z
1 L{vo (t)} 1
3) vo (t) = i1 (t)dt = R2 i2 (t) −−−−−→ Vo (s) = I1 (s) = R2 I2 (s)
C Cs
From the above equations, we obtain
1
Eo (s) I (s)
Cs 1
G(s) = = 1
Ein (s) R1 I(s) + Cs I1 (s)
1
I (s)
Cs 1
= 1
R1 [I1 (s) + I2 (s)] + I (s)
Cs 1
1
= R1
(8.1)
R1 Cs + [1 + R 2
]
As in Eq.(8.1), the transfer function Eo /Ein is not only dependent on R1 , C but also
dependent of R2 , which indicates that the plant composed of R1 and C is not an isolated
system. In other words, it is affceted by the next adjacent part of circuit which is called
a load of the previous circuit. In conclusion, it is unavoidable to have a non-isolated
system if a system is passive circuit. In order to have an isolated system, a system
should be feedback system(servo system) such as an operational amplifers.

2 The load effect modeled using an electronic circuit


A load is defined as any kinds of components that reduce the magnitudes of effort sources
or flow sources. They can be easily understood at the bond graph. As an example, let
us figure out how loads reduce the applied voltage using the electrical circuit shown in
Fig. 7.8 represented using the bond graph with an effort source applied. The constant
voltage input is reduced as it is encountered with several loads such as R1 , C, and R2 .
Hence, e2 is smaller than ein by the amount of eR1 . The current i3 is also reduced by
the amount of iC . The voltage and current reduced at each load can be obtained using
the constitutive relations associated with loads, R1 and C.

There is another way to calculate the effort and flow variables changed due to the
load effects using an electronic circuit model for general systems. For example, when a
load is modeled using a resistor as shown in Fig. 8.2, the current is obtained form the
electical relation of iR = 10V
1k
when the applied voltage is assumed to be constant 10 V .

168
Figure 8.2: A load modeled using a resistor

When another load is added in series or pararell to the inital circuit as shown in Fig.
8.3, it is unfortunate not to use the analysis done for the initial curcuit. A new analysis
should be done for a new circuit to calculate an effort or a flow variables because the
system is passive.

Figure 8.3: An electric cicuit with another load added in parallel

169
Similary, this load effect is also found in a mechanical system. When a load is applied
to the initial load in series as shown in Fig.8.4, a new analysis for the 2 DOF mechanical
system should be done from the beginning for the new system.

Figure 8.4: Mechanical system with another load added in series

One more thing to note here regarding to the load effect is the effort or flow source
can be also changed due to the load effect as well. Even though the voltage source was
assumed not to be changed in the above statement, the voltage can be also changed if
unexpected loads are added to the sources in reality. In a hydraulic system, for example,
the pressure source Pc can be decreased when there is a large flow across the nozzle shown
in Fig. 8.5 and the decreased pressure is calculated using the load condition and constant
power curve. Depending the load condition, Pc = Rn Q and the constant power curve
Pc Qc = constant, the input pressure and flow can be found out.

170
Figure 8.5: Hydraulic system (a) modeled using an electronic circuit and (b) bond graph,
(c) load curve

3 Operational amplifier free from a load effect


An operational amplifier is one of typical examples which takes an advantage of no
load effect using an electronic feedback system. It has characteristics of a high input
impedance and low output impedance.

For a reference, the impedance of a device Z is defined as ei . Let’s look at a simple


electronic circuit composed of a resistor R and capacitor C shown in Fig. 8.6(a).

Figure 8.6: A simple electronic circuit composed of a resistor R and capacitorC

Figire 8.6(b) shows its bond graph model showing all components and 8.6(c) shows an
equivalent bond graph modeled using just an equivalent resistor Req . The equation of

171
the electronic circuit is represented
Z
1 1
e = Ri + idt = (R + )I (8.2)
C Cs
Hence, the impedance of the resitor Z is is obtained as
e 1
Z= =R+ (8.3)
i Cs
As represented in Eq.(8.3), Z is depedndent on frequency.

Since the operational amplifier has a high input impedance, it prevents the current
from flowing into the next circuit represented with a red dotted line in Fig. 8.7. That is
why the previous circuit can be considered as an isolated circuit from the next circuit.
The low ouput impedance condition can flow the current from the power supply well
through the operational amplifer to the next circuit.

Figure 8.7: Hihg input impedance and low output impedance device

3.1 Voltage amplifier

Figure 8.8: An operational voltage amplifier

172
Figure 8.8 shows a schematic diagram of an operational voltage amplifier. Due to the
condition of high input impedance of the op amplifier, the input current iin can not
flow through the resistor but flow through the feedback resistor Rf . In other words, the
feedback loop makes the output current if equal the input cuurent iin . These conditions
lead to the following resuts.

v+ = v− = 0 (8.4)
vi − v − vi
iin = = (8.5)
R1 R1

v − vo vo
if = =− (8.6)
Rf Rf
iin = if (8.7)

From the above equations, we have

Rf
vo = − vin (8.8)
R1
It is possible to make several isolating systems using op amplifers having unit gains(or
buffers) between the passive systems as shown in Fig. 8.9. Hence, there is no need to
do circuit anaysis again when a new passive circuit is added to the intial circuit. It is a
great advantage of using operatiomal amplifiers.

Figure 8.9: Op amplifers between the passive systems to make isolating systems

The name of the voltage amplifer comes form the concept of the feedback of the
output voltage vo as the name of the position controller(amplifier) comse frome the
postion feedback(vo is equal to -vf ). This relation is proven as below

v1 Rf
vf = Rf if = Rf = v1 = −vo (8.9)
R1 R1

173
Recalling the ideal condition of the voltage source in section 3.2 of Chapter 1, the op
amplifier is consdiered as a voltage source because it has a low output impedance, that
is, no voltage loss from an internal resistor. Therefore, when a load RL is connected
to the ouput voltage, the current is determined from the relation of RVoL and it is drawn
from a power supply.

Q: Figure 8.10 shows a non-inverted amplifier. Take steps of the circuit


analsys to obtain the input and output relation

Figure 8.10: A non-inverted amplifier

3.2 Ideal voltage source

Figure 8.11: Ideal voltage source

An ideal voltage source is defined as a voltage supply that provides the input voltage to a
load without a voltage loss as shown in Fig8.11. Therefore, the internal serial resistance
Rin should small enough to be ignored, that is, Vi n = 0 so that V is fully applied to
RL . Here, note that the load voltage VL is measured using a oscilloscope that has an
∞ impedance in order not affect the load RL . If the impedance of the oscilloscope is

174
similar to RL , VL would be smaller than V , which is not intented to be.

Q: What is the voltage across RL = 1kΩ when the voltage source 10 V is


applied to the circuit shown Fig. 8.11 for Rin = 50Ω? How is the the voltage
across RL = 1kΩ changed when Rin = 1kΩ?

Q: What is the voltage across RL = 1kΩ measured by the osciloscope having


the resistance of 1M Ω at the circuit shown Fig. 8.11 for Rin = 0 ? What is
the voltage across RL = 100kΩ measured by the osciloscope having the same
resistance?

3.3 Current amplifier

Figure 8.12: An operational current amplifier

Figure 8.12 shows a current amplifier used for amplifying the current proportionally to
the input voltage. There is a feedback loop connecting the point having Va to the point
having v . Due to the condition of high input impedance of the op amplifier, the input
current iin can not flow through the resistor of the op amplifier but flow through the
feedback resistor Rf . In other words, the feedback loop makes the output current if
equal the input cuurent iin . These conditions lead to the following resuts. The load
current IL is divided into the feedback current if and Is . The above conditions lead to

175
the following results

v+ = v− = 0 (8.10)
vi − v − vin
iin = = (8.11)
R1 R1
vs
if = = −iin (8.12)
Rf
 
vi Rf
vs = Rf if = Rf − = − vi (8.13)
R1 R1
 
vs 1 1
iL = if + is = + vs Rs = vs + vi (8.14)
Rf Rf Rs
 
vin Rf
vs = Rf if = Rf − = − vin (8.15)
R1 R1

From the above equations, we have the relation that the ouput current IL is amplified
by the input voltage vin as
 
Rf 1 1 Rf (Rs + Rf ) Rs + Rf
iL = − + vin = − vin = − vi (8.16)
R1 Rf Rs R1 Rf Rs R1 Rs

The name of the current amplifer comes form the concept of the feedback of the
output current iL though the actual feedback current is the partial of it. Recalling the
ideal condition of the current source in section 3.4 of Chapter 1, the current amplifier is
consdiered as a current source because it has a high input impedance, that is, no current
loss from an internal resistor. Therefore, when a load RL is connected to the ouput,
the current is determined by the input voltage regardless of the magnitude of the load
resistor and it is drawn from a power supply. Here, RL is supposed to be smaller than
the input impedance of the op amplifier.

Rs is used to flow the current to the direcion to the ground because Rf is usually
chosen to be much larger than RL . Hence, Is ∼= IL . If Rs does not exist, the high current
output will flow to Rf , which will burn out the op amplifier. Rs is also used to monitor
IL in terms of voltage. The power consumption in Rs is calculated as Is2 Rs . So, the
material property of Rs should be strong enough not to be burnt.

3.4 Ideal current source


An ideal current source is defined as a current supply that applies the input current to a
load without having a current loss as shown in Fig. 8.13. Therefore, the internal parallel

176
Figure 8.13: Ideal current source

resistance Rin should be ∞ enough to be treated an open circuit, that is, Iin = 0 so that
I is fully applied to RL . Here, the load current IL is measured using a current meter
that has an 0 impedance.

4 Filters using op amplifiers


We obtained the transfer function of an amplifer by deriving the electronic equations
associated with the circuit. They are also easily obtainable by the ratio of the impedance
of the feedback loop Zf to that of the feed-forward loop Zi . Thus,
Zf
G(s) = − (8.17)
Zi
The low pass filter can be constructed as shown in Fig. 8.14.

Figure 8.14: A low pass filter using an op amplifier

The transfer function of the low pass filter is obtined as


1
1/R2 +(Cs) R2
G(s) = − =− (8.18)
R1 R1 (R2 Cs + 1)

177
Q: Check if Eq.(8.18) is also obtained using the circuit analysis.

The high pass filter can be constructed as shown in Fig. 8.15.

Figure 8.15: A high pass filter using an op amplifier

Then, the transfer function of the high pass filter is obtined using Eq.(8.17) as

R2 C1 s
G(s) = − (8.19)
1 + R1 C1 s

Q: Check if Eq.(8.19) is also obtained using the circuit analysis.

Various filters used for controllers are listed in Fig. 8.16.

178
Figure 8.16: Various filters made of operational amplifiers

179
5 Controller design using op amplifiers
5.1 Proportional Controller
Figure 8.17 shows an integrator made using an op amplifier whose transfer funtion is
1
RCs
. When an input applied to the integrator, the output is obtained by integrating the
input. For example, I V DC input is appled, the output is t. When cos ωt is applied,
the output is sin ωt

Figure 8.17: An integrator usign an operational amplifier

This transfer function can be also found at the mechanical model in which force is
an input to a mass m and velocity is an output as shown in Fig.8.18. This is one of
popular mechanical systems.

Figure 8.18: Integrator in a mechanical system

Since the dynamic equation of the mechanical system shown is F = m dv


dt
, we have the
open loop transfer function G(s)

V (s) 1
G(s) = = (8.20)
F (s) ms

When a 1N force is appied to the mass 1kg, the velocity of the mass is obtained by
integrating the input force, hence we have the velocity v = t, v(t) increases proportion-
ally with respect to time. If we know the parameter accurately we can manage F to

180
have a desired velocity using the mathematical relation. However, if we can not know
the paramater correctly it is not easy to expect the output behavior correctly. More-
over, it is difficult to calculate the force when zero velocity is required during the motion.

A feedback loop is constructed to have a constant velocity. Figure 8.19 shows a


standard form of a feedback system represented using a block diagram of the open loop
transfer function G(s), feedback transfer function H(s), and a controller transfer func-
tion C(s) having a gain K. The output velocity is fedback to the input so that the
velocity error e is decreased. For example, when the output velocity is is bigger than
the desired velocity 0, a negative force is applied so as to reduce the velocity. When the
output velocity is smaller than the desired velocity 0, a positive force is applied o as to
increase the velocity.

Figure 8.19: A feedback system represented using a block diagram and a proportional
controller P

From Fig.8.19, we have the relaions

E(s) = Vd (s) − V (s)H(s)


F (s) = C(s)E(s)
V (s) = G(s)F (s)
V (s)
hence, the closed loop tranfer function Gcl defined as Vd (s)
is obtained as
V (s) C(s)G(s)
= (8.21)
Vd (s) 1 + C(s)G(s)H(s)
Here. the controller C(s) is called a proprotional controller since the gain is contant K.
Then, using the information C(s), G(s), and H(s) = 1, we obtain
K 1
Gcl = = m (8.22)
K + ms 1+ K s

181
What is the difference between G(s) and Gcl ? The integrator represnted in G is changed
to a low pass filter in Gcl .

Q: How should the controller gain be adjusted to have a large cuf-off frqeuncy
in the closed loop system. Sketch the time domain response v(t) when the
input vd is 1m/sec and 1 sin(10t)m/sec when the mass is 1kg and K is 10. Compare
the results with the open loop results

5.2 PD controller

Figure 8.20: A mechanical system composed of mass and damper

What will happen when there is a damper added to a mass in a mechanical system
as showin in Fig.8.20? We are interested in position this time. The dynamic equation
corrseponding to the mechanical system is F = m dv dt
+ bv, v = dx
dt
, henec, we have the
open loop transfer function G(s) with input force and output position as

X(s) 1
G(s) = = (8.23)
F (s) s(ms + b)

When the proportional controller is replaced by a PD(proportional and derivative)


controller described by C(s) = Kp (1 + Td s) where, Kp and Td are the proportional gain
and derivative gain, respectively, the closed loop trnasfer fuction Gcl is obtained using
Eq.(8.21),
C(s)G(s) Kp (1 + Td s)
Gcl = = 2
(8.24)
1 + C(s)G(s)H(s) ms + (b + Kp Td )s + Kp )
The block diagram of the closed loop can be represented as shown in Fig. 8.21.

182
Figure 8.21: The block diagram of the closed loop system with PD controller

Equation(8.24) indicates that the integrator plus the first ordered low pass filter sys-
tem is converted to a second ordered low pass filter system having the natural frequency
and dmaping ratio in the closed system. In other words, a mass-damper system in the
open loop showin in Fig. 8.22(a) chagned to a mass-electric spring-electric damper in
the closed loop system as shown in Fig. 8.22(b). The spring and damping effect can
be adjusted using the gains, Kp and Td so as to give a desired natural frequency and
dmaping ratio. One thing to note here is the marginally stable system composed of the
mass and damper is stablized by the feedback system that is equivalently composed of
the mass, spring, and damper, which is one of great advantages of the feedback system.

Figure 8.22: (a) The integrator plus the first ordered low pass filter system in an open
loop and (b) a second ordered low pass filter system uisng the PD controller.

Please refer to Chapter. 6 for more detail to study the stability analysis of the closed
loop system in a frequency domain

Q: What is the equivalent electrical circuit model corresponding to Eq.(8.23)


? What are the input and output?

Q: Please find out experimentally how the output magnitude is respectively


changed for a low freqeuncy and high frequency inputs

183
5.3 Stability
A plant is called stable when the roots of the characteristic equations of the plant,
namely, poles are located in the left half s plane. The plants described in Eq.(8.20) and
Eq.(8.23) are marginally stable because the poles s = 0 are not located in the left half
plane of s as shown in Fig. 8.23. When the closed loop Gcl is realized using the PD
controller and the plant respresented in Eq. (8.24), the characteristic equation is

ms2 + (b + Kp Td )s + Kp ) = 0 (8.25)

Then, the poles are obtained using the standard form of s2 + 2ζωn s + ωn 2 = 0, where
ωn 2 = Kp , 2ζωn = b + Kp Td . We can observe that the poles of the closed loop are
changed to s = −σ ± ζωn with all poles in the left half s plane. It can be said that the
PD controller makes the plant have better performances in terms of stability and speed.

Figure 8.23: Poles of the closed loop and open loop systems.

6 Controller design for a real mechanical system


So far, the plants were implemented using electronic circuits, an integrator, first order
low pass filter, and second order low pass filter. In reality, they are real systems made
of mechanical structures. When this real system is considered for control, the control

184
configuration has different components.

Figure 8.24 shows a real configuration for control of a mechanical system with the
consideration of the variables of the inputs and outputs of the components. Here,
ed , e, u, i, F, x, and ex are the command input, error, input to the amplifier, current,
force, position, and output. The amplifier is required to supply a high voltage or current
to the actuator such as an electrical motor. Force is generated from the motor based
on the electromechanical principle. The sensor is also required to measure the ouput
variable such as position and it is fed to the command input to generate the position
error. Since the real plant was simulated using the analog circuit, it was not necessary
to have the amplifier, actuator, and sensor in the previous feedback systems.

Figure 8.24: The block diagram of the closed loop system for a real application

For example, when a steel ball is levitated using the magnetic force as shown in Fig.
8.25, the mechanical equation is

Fem − mg = mẍ (8.26)

Here, Fem is represented by the varaibles i and x and its relation is experimentally ob-
tained and it is shown in Fig. 8.26

185
Figure 8.25: A steel ball levitated using the magnetic force

Figure 8.26: The electromagnetic force varied with respect to current and position

Fem can be represented at the operating position or equibrium position xo , and io as


Fem = K1 x + K2 i (8.27)
K1 and K2 can be respectively obtained by

δFem
K1 = |i=io (8.28)
δx
δFem
K2 = |x=xo (8.29)
δi
186
Then, using using Eq. (8.26) and (8.27), we have

K2 i(s) = (ms2 − K1 )X(s) − mg (8.30)


Since i can be considered as the summation of a constant current I for holding the steel
ball in the equibrium position and applied current i for controlling the ball, Eq.(8.30) is
modified as
K2 i(s) = (ms2 − K1 )X(s) (8.31)
Hence, when i and x are the input and output, the transfer function G(s) is

X K2
G(s) = (s) = 2
(8.32)
i (ms − K1 )

The transfer function represented in Eq. (8.32) has poles in the right half plane,
which is unstable. Therefore, control is required to make the plant stable. Figure 8.24
is modified to represent the control block diagram of the levitated ball as shown in Fig.
8.27.

Figure 8.27: The control block diagram of the levitated ball

Q: What kinds of a controller can be used to make the levitated ball stable
in the feedback system? Prove the feedback system is stable based on the
pole location of the closed loop system.

1
Q: Consider a mass-damper plant whose transfer fuction is ms+b . Here, the
force is an input and velocity is an output. The mass is 0.1 kg and damping
coefficent is 20 N · sec/m. The plant is implemented using an electronic cir-
cuitry composed of operational amplifiers having the same transfer function
as shown in Fig. 8.14 . Use C2 = 0.1µF.

187
1. Design a PD controller Kp (1 + Td s) referring to Fig. 8.16 for velocity
control. How the control gains affect the system performances in terms
of speed. Compare the reference input and velocity output signals when
the reference inputs are a unit step and ramp input. Make theoretical
analysis of the result.

2. When a PI controller Kp (1 + T1d s ) is used for the same plant, compare the
reference input and velocity output signals when the reference inputs
are a unit step and ramp input. Make theoretical analysis of the result

188

You might also like