Beruflich Dokumente
Kultur Dokumente
Peter Grogono
December 2003
Contents
1 Syllabus
2 Transforms
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
5
9
10
11
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5 Stability Criteria
5.1
5.2
5.3
5.4
Routh-Hurwitz
Nyquist . . . .
Bode . . . . . .
Root-Locus . .
25
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6.1
6.2
6.3
6.4
Zero Input . . .
Stability . . . .
Compensation .
Normal Form .
13
14
15
16
17
17
23
24
24
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
26
28
29
30
32
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
34
35
35
36
7 SI Units
37
8 Useful Formulas
41
List of Figures
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
5
7
8
11
12
14
15
15
18
18
19
22
24
30
32
37
38
41
42
44
45
List of Examples
1
2
3
4
5
6
7
8
9
10
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
6
7
10
10
11
20
20
20
21
SYLLABUS
Syllabus
Transforms
2.1
A periodic function f(t) with period T can be expressed as a sum of sine and cosine functions:
f(t) =
2X
a0
+
(an cos n t + bn sin n t)
T
T
n=1
2n
.
T
n = 1, 2, 3, . . .
Using the formulas cos = 21 (ej + ej ) and sin = 2j (ej ej ), we can express this sum using
complex exponents as
f(t) =
i
a0
1 Xh
(an jbn )ejn t + (an + jbn )ejn t
+
T
T
n=1
where
Z T/2
f(t)ejn t dt
an jbn =
T/2
Z T/2
f(t)ejn t dt
an + jbn =
T/2
(1)
TRANSFORMS
Z
T/2
1 X
f(t)ejn t dt ejn t .
T n= T/2
1 X
Cn ejn t
T n=
f(t) =
where
Z T/2
f(t)ejn t dt.
Cn =
T/2
and assume that f(t) is this function extended periodically with period T | in other words, f is a
pulse train. (Clearly, we require p < 2T .) Then
Z p/2
Aejn t dt
Cn =
p/2
A jn t p/2
e
jn
p/2
2A
p
sin n
n
2
=
=
and
f(t) =
1 X 2A
n p jn t
sin
e
T
n
2
n=0
AX1
n p jn t
sin
e
n
2
n=0
since
2
1
=
.
Tn
n
F() =
f(t)ejt dt
which is the Fourier transform of an arbitrary (i.e., non-periodic) function of time, f(t).
The Fourier transform has an important limitation: it is valid for a function f(t) only if
|f(t)|dt
has a nite value. For example, the \step function", u(t), has no Fourier transform. The Laplace
transform introduces an additional \damping factor", est , and consequently can be used for a
greater variety of functions.
TRANSFORMS
f(t)
6
p -
p -
- t
p
2
2T
L (A f(t) + B g(t)) =
= A L (f(t)) + B L (g(t)) .
The transform of a derivative has a simple relationship to the transform of the original function.
Using the formula for integration by parts
Z
uv =
(2)
u dv + v du
st
f(t) e
Z
= s
f(t) e
0
st
Z
dt +
L f 0 (t)
= s L (f(t)) f(0+ ).
(3)
TRANSFORMS
Dierentiating again:
L f 00 (t)
= sL f 0 (t) f 0 (0+ )
f() d
u(t) =
0
st
v(t) = e
in (2) gives
Z Zt
Zt
est f() d
0
f() d
= s
0
st
dt +
f() d
R0
0
1
L (f(t)) .
s
Suppose that f(t) is a variable in a control system. Then typical operations on f scale it by
a linear factor (e.g., amplier), dierentiate it (e.g., inductor), or integrate it (e.g., capacitor).
Consequently, the Laplace transform of f is likely to be a rational function of s or, in other words,
P(s)
in which P(s) and Q(s) are polynomials in s. The
it will be a function of the form M(s) =
Q(s)
roots of P are the zeroes of M and the roots of Q are the poles of M.
Theorem 2.1 Initial Value
t0
s0
0
e(a+s)t
a + s 0
1
= 0
a+s
1
=
.
s+a
TRANSFORMS
Function
Transform
Remarks
f(t)
F(s)
General notation
Af(t) + Bg(t)
AF(s) + BG(s)
Linearity
f(t T )u(t T )
esT F(s)
eat f(t)
F(s + a)
f 0 (t)
sF(s) f(0+ )
f 00 (t)
Rt
0 f()d
Rt
0 f(t )g()d
1 est+jt
1 estjt
+
2 s + j 0
2 s j 0
1
1
1
+
2 s + j s j
1 s j s + j
=
+
2 s2 + 2
s 2 + 2
s
= 2
.
s + 2
Let
u = sin t
du = cos t dt
v = est
dv = sest dt
R
st
Z
= s
0
sin(t)e
st
Z
dt +
0
cos(t)est dt
TRANSFORMS
Function
Transform
Remarks
(t)
Dirac's -function
1
s
1
sn
1
s+a
1
s(s + a)
s
2
s + a2
s
s2 a2
a
2
s + a2
a
2
s a2
1
s(s2 + a2 )
1
s2 (s2 + a2 )
1
(s + a)2
s
(s + a)2
1
2
s + 2 + 2
u(t)
tn1
(n 1)!
eat
1 eat
a
cos at
cosh at
sin at
sinh at
1 cos at
a2
at sin at
a3
teat
eat (1 at)
eat sin bt
b
a = , b = 1 2 , and < 1
p
and therefore
L (sin t) =
=
2
L (cos t)
s
.
s2 + 2
TRANSFORMS
2.2.1
Partial Fractions
The last step of a solution that employs Laplace transforms usually requires the inverse transforP(s)
mation of an expression of the form
in which P and Q are polynomials in s. Since the inverse
Q(s)
1
transform of
is straightforward, we need to express such rational functions in the form
s+a
A2
A3
A1
+
+
+
s + s1 s + s2 s + s3
then
P(s)
Q(s)
n
X
Ai
s + si
i=1
where
Ai =
(s + si )
P(s)
.
Q(s) s=si
The situation is slightly more complicated when there are repeated roots. If
Q(s) = (s + si )r
P(s)
will include the terms
Q(s)
B1
B2
Br
+
+ +
2
s + si (s + si )
(s + si )r
in which
Br =
Br1 =
Br2 =
..
.
B1 =
(s + si )r
P(s)
(s + si )
Q(s) s=si
P(s)
1 d2
(s + si )r
2
2! ds
Q(s) s=s
1 d
1! ds
P(s)
Q(s) s=si
r
dr1
1
(r 1)! dsr1
r P(s)
(s + si )
Q(s) s=s
10
TRANSFORMS
2.3
The Z-Transform
A sampling system takes \samples" of a continuous input at discrete points in time. Each
sample can be represented by Dirac's function with an appropriate amplitude. Thus if e(t) is a
continuous signal sampled at times 0, T, 2T, 3T, . . ., the sampled signal is given by
e (t) =
e(nT )(t nT )
n=0
E (s) =
e(nT )enTs .
(4)
n=0
In order to avoid the non-algebraic terms that would be introduced by enTs , we change the
variables by dening
z = eTs .
ln z
T
, enTs = eTs
n
E (s) = E
ln z
e(nT )zn
n=0
= E(z)
and we say that E(z) is the z-transform of e(t). Figure 4 shows some common z-transforms.
In general, E(z) = e(0) z0 + e(T ) z1 + e(2T ) z2 + + e(nT ) zn + and we can think of the nth
coecient as representing the function at time nT .
Example 4: z-transform of unit step. For the unit step function, r(t) = 1 for t 0 and
therefore e(nT ) = 1 for all n 0. Thus
R(z) =
zn
n=0
z
.
z1
2
Example 5: z-transform of unit ramp. For the unit ramp function, r(t) = t for t 0 and
therefore e(nT ) = nT for n 0.
R(z) =
nTzn
n=0
= T (z1 + 2z2 + )
z
= T
.
(z 1)2
11
Remark
(t)
Impulse
z
z1
z
z1
Tz
(z 1)2
T 2 z(z + 1)
2(z 1)3
z
z eaT
(1 eaT )z
(z 1)(z eaT )
z sin T
2
z 2z cos t + 1
Unit step
u(t)
T (t)
t
1 2
t
2
eat
1 eat
sin t
T (t) =
(t nT )
n=0
R(z) =
eaTn zn .
n=0
We note that
R(z)
R(z)zeaT
=
= zeaT
+1 + eaT z1 + e2aT z2 +
+1 + eaT z1 + e2aT z2 +
and therefore
R(z)zeaT R(z) = zeaT
from which
R(z) =
=
zeaT
zeaT 1
z
.
z eaT
12
All functions are expressed as Laplace transforms of the corresponding time-domain functions.
E(s)
R(s)
6
G(s)
C(s)
B(s)
H(s)
and therefore
C(s) =
R(s) G(s)
1 + G(s) H(s)
C(s)
G(s)
=
.
R(s)
1 + G(s) H(s)
Since the transfer function without feedback is simply G(s), the denominator 1 + G(s)H(s) represents the eect of feedback. Note that the sign of the feedback (negative) is assumed: B(s) is
subtracted from R(s).
The characteristic equation is obtained by equating the denominator of the transfer function
to zero. For the system in Figure 5, the characteristic equation is R(s) = 0. In many cases | for
example, when G(s) is a simple polynomial | it will be
1 + G(s)H(s) = 0.
13
From the form of the characteristic equation, we note that G and H have reciprocal units.
In the following example, the units of G are radians/volt and the units of H must therefore be
volts/radian.
Kirchoff s Laws
Current Law (KCL). The algebraic sum of the currents leaving/entering a node is zero.
Voltage Law (KVL). The algebraic sum of the potential dierences around a loop is equal to
the algebraic sum of the e.m.f.'s acting around the loop in the same direction.
4.1
the relationship between phasors (see below) when voltage and current vary sinusoidally; and
the Laplace transform of the time function.
Resistance. For resistance, the time function (and hence the other functions) is very simple:
v(t) = R i(t).
Capacitance. A capacitance stores charge, giving
i = C
or
v =
1
C
dv
dt
Zt
i d.
=0
di
.
dt
Component
Time
Frequency
Laplace
v = f(i)
V()/I()
V(s)/I(s)
Ri
1/jC
1/sC
jL
sL
Resistance
Capacitance
Inductance
1
C
i dt
di
dt
14
v = L
= jLi
and therefore
v
i
= jL
in which jL is a phasor that expresses the impedance of the inductance. In general, we can
forget about the term ejt in our calculations, because it cancels out in all the equations. We can
ignore that amplitude, A, for the same reason.
The three forms are summarized in Figure 6.
It is useful to note that there are several ways of combining components to give terms of particular
dimensions. For example, RC has the dimensions of time and RC is therefore a pure number.
Similarly, LC has dimension sec2 and LC is therefore a pure number. See Section 7 for more
examples.
4.2
Phase Control
Figure 7 is a general circuit for phase control. If i is the current
owing through Z1 and Z2 , we
have
Vin = i(Z1 + Z2 )
Vout = iZ2
Z2
.
Z1 + Z2
15
Vin
Z1
Vout
Z2
Lead Compensation
s+a
= tan1
tan1
a
b
= arg
16
Z1 =
Z2
The response is
Vo
Vi
Z1
Z1 + Z2
R2 (1 + R1 Cs)
R1 + R2 + R1 R2 Cs
s + R11C
=
=
=
s + R11C +
s+a
s+b
1
R2 C
Lag Compensation
s+b
= tan1
tan1
b
a
= arg
1
Cs
1
Cs
R2 + 1/Cs
R1 + R2 + 1/Cs
Z1 + Z2 = R1 + R2 +
Z2
Z1 + Z2
17
s + 1/R2 C
s + 1/(R1 + R2 )C
s+b
s+a
=
=
Lag/Lead Compensation
(s + a1 )(s + b2 )
with a1 < b1 and a2 < b2 .
(s + b1 )(s + a2 )
Z1 =
Z2
Z2
Z1 + Z2
s2 +
s+
1
R2 C 2
1
R1 C1
1
R2 C1
s+
1
R2 C 2
1
R1 C 1
s+
1
R1 C 1 R2 C 2
Table 9 summarizes the advantages and disadvantages of the various kinds of compensation.
4.3
Operational Amplifiers
Figure 10 shows the circuit symbol for an operational amplier (\op amp") and the equivalent
circuit. In designing with op amps, we assume:
The input impedance is large (Zin )
The output impedance is small (Zout 0)
18
Compensation
Advantages
Phase lag
LF characteristics improved
Phase lead
Disadvantages
at least one slow term in transient reponse
reduced bandwidth may be a disadvantage
possible HF noise problems
may generate large signals, out of
linear range of system
Figure 11 shows op amps in two typical congurations. Each circuit uses two impedances, Z1 and
Z2 , in a feedback loop. The input signal for the left circuit (a) goes to the input of the op amp,
and the circuit has negative gain. The input signal for the right circuit (b) goes to the + input of
the op amp, and the circuit has positive gain.
To analyze Figure 11(a), we assume that the input impedance of the op amp is high enough to
ensure that the current
owing into its input is negligible. Consequently, we can assume that
the same current
ows through Z1 and Z2 , and we will call it i. It follows that:
v Vin = i Z1
Vout v = i Z2
Vout = A v
Vin
Zout
Zin
Vin 6
?
Z
+
A Vin
Vout
Vout
19
Z2
Vout
v
Vin
Z1
Z2
Vout
Vin
Z
Z
Z1
Vout v
Z2
and therefore
Vout
Vin
Z2
.
Z1 + Z2
Z1 +
A
Z2
.
Z1
We can analyze Figure 11(b) in a similar way. Again, we assume that the currents
owing into the
inputs of the op amps are negligible and that the current
owing through Z1 and Z2 is i. Thus
v = i Z1
Vout = i(Z1 + Z2 )
and, eliminating i,
v
Z1
Vo ut
Z1 + Z2
or
v =
Vout Z1
.
Z1 + Z2
20
Z1
1
.
Z1 + Z2 A
Z1
.
Z1 + Z2
Z2
Z1
Z1
Z1 + Z2
R2
R1
and the circuit behaves as an (inverting) linear amplier with gain R2 /R1 . 2
Example 8: Differentiation. If Z1 = 1/Cs is a capacitor and Z2 = R is a resistor,
T (s) = RCs
R
.
1 + RCs
R1
1 + R1 C1 s
R2
1 + R2 C2 s
Z2
Z1
R2
1 + R1 C1 s
1 + R2 C2 s
R1
R2 1 + R1 C1 s
R1 1 + R2 C2 s
For a d.c. signal, s = j = 0 and T (s) = R2 /R1 , as we would expect. For a.c. signals, the gain is
g
R2 1 + R1 C1 j
=
R1 1 + R2 C2 j
v
u
u
R2
= t
R22 C1 C2 2
+
R1 (1 + R22 C22 2 ) 1 + R22 C22 2
!2
R22 C2
R2 C1
!2
21
2
Example 10: Simulating an inductance.
Figure 12 shows a circuit that is intended to simulate an inductance. To demonstrate its action, we
compute the eective impedance between the two input connections at the left of the diagram. To
obtain the impedance, we apply a voltage V to these terminals and calculate the resulting current.
Let i1 be the current
owing through R1 and i2 be the current
owing through C and R2 . We have
u = V + i1 R1 ,
i2
,
v = V+
sC
i2
.
V = i 2 R2 +
sC
(5)
(6)
(7)
The eect of the op amp is to keep the voltages at points u and v approximately equal. Assuming
V + i1 R1 = V +
i2
sC
and therefore
i2
.
sR1 C
i1 =
From (7),
i2 =
V sC
1 + sR2 C
and so
i1 + i2 =
=
V sC + V s2 R1 C2
sR1 C(1 + sR2 C)
V sC 1 + sR1 C
sR1 C 1 + sR2 C
Z =
Z =
c
6
c
22
R1
Z
R2
1
R2 R1
(8)
(9)
(10)
In these circumstances,
R 0 R2
L0
Q
R2
2 R1 C
1
R1 C
It appears from these approximations that we can obtain large values of both L 0 and Q by using
a very small value of C. Small values of C, however, may invalidate the inequalities (8){(10).
In principle, we can get some quite impressive inductors. For example, suppose that we use the
values R1 = 10 K, R2 = 1 M, and C = 1 nF. At f = 1 KHz, the values of R 0 , L 0 , and Q are shown
below. The values given by the approximate formulas, shown in parentheses, are not accurate at
this frequency.
R 0 = 13, 893 (1 M)
L 0 = 9.86 H
(2533 H)
Q = 4.46
(15.9)
23
These results suggest that when the approximations are useful, the inductor is not useful, and vice
versa. C'est la vie. 2
4.4
Mechanical Systems
Equations for mechanical systems enable us to calculate the motion caused by a force.
A mass obeys Newton's law: force = mass acceleration or
d2 x
dt2
dv
= m .
dt
f(t) = m
In an electrical system in which e is the potential, q is the charge, i is the current, and L is an
inductance, we have the analogous equations
d2 q
dt2
di
= L .
dt
e(t) = L
q
.
C
= v
= i.
Friction is non-linear in general. However, we can approximate viscous friction as a force propor-
tional to velocity:
f(t) = v
Rotating systems are similar but, instead of mass, force, and velocity, we have torque, moment
of inertia, and angular velocity. Figure 13 summarizes the analogies between mechanical and
electrical systems.
24
Linear
Symbol
Rotating
Symbol
Electrical
Symbol
Position
Angle
Charge
Mass
Moment of Inertia
Inductance
Velocity
v=
dx
dt
Angular Velocity
Torque
T = I ddt2
Potential
e = L ddt2q
C de
dt = i
d
dt
Current
Force
f = m ddt2x
Spring
k df
dt = v
Clockspring
dT
dt =
Capacitance
Friction
v = f
Angular friction
= T
Resistance
i=
dq
dt
2
Ri = e
Quality Factor
The quality factor, Q, of a resonant circuit is a measure of the height of the \spike" in the
response function. The Q of an LC circuit is typically determined largely by the resistance, R, of
the inductor L, and
Q =
L
.
R
Quality does not apply only to electronic circuits, however, and the general denition of Q is given
in terms of the bandwidth of the \spike".
Definition 4.3 Quality Factor Let S be a resonant system with resonant frequency (that is,
maximum response) at 0 . Let 1 and 2 , where 1 < 0 < 2 , be the \half-power points",
where the amplitude is 3 dB (or 1/ 2) down. Then the quality factor for S is
Q =
0
.
2 1
mechanical oscillator, where and are linear and angular friction as described in Figure 13, and
4.6
Second-Order Systems
2n
.
s2 + 2n s + 2n
25
STABILITY CRITERIA
2n
s(s2 + 2n s + 2n )
= 1
1 n t
e
sin(n t + )
where
q
1 2
= tan1 (/)
12 .
Stability Criteria
Definition 5.1 Stability A system is stable if, for every bounded input, the output remains
bounded with increasing time.
In most case, G(s) and H(s) are rational functions of s and we can write
1 + G(s)H(s) =
(s z1 )(s z2 )(s z3 )
(s p1 )(s p2 )(s p3 )
(11)
in which the factors in the numerator are (mis)named the zeroes and the factors in the denominator are (mis)named the poles of the CE.
Recalling that the closed-loop transfer function is
C(s)
R(s)
G(s)
1 + G(s)H(s)
R(s)G(s)
1 + G(s)H(s)
k2
kn
k1
+
+ +
+ Cr (s)
s p1 s p2
s pn
and consequently
c(t) = k1 ep1 t + k2 ep2 t + + kn epn t + cr (t)
(12)
26
STABILITY CRITERIA
Since the transfer function is a polynomial, cr (t) will be bounded for bounded inputs c(t). The
other terms will be bounded if all of the pi are negative. If there are quadratic factors, some of
the pi will be complex, and must have negative real parts. It follows that, for stability, the roots
of the CE must have negative real parts. If the CE has roots on the imaginary axis, the system
is marginally stable and, in practice, will oscillate.
Consequently, the stability analysis depends on being able to nd the zeros of a polynomial. We
begin with some general considerations.
The equation (s r1 )(s r2 ) (s rn ) = 0 has roots at s = r1 , s = r2 , . . . s = rn and expands to
sn + an1 sn1 + an2 sn2 + + a0 = 0.
We have:
an1 = sum of all roots
an2 = +sum of products of pairs of roots
an3 = sum of products of triples of roots
...
a0 = (1)n product of all roots
If all of the roots real and negative, then ri < 0 for i = 1, 2, . . . , n and it follows that aj > 0 for
j = n 1, n 2, . . . , 0. If the roots are complex, they must occur in conjugate pairs. Since the ai
are real, the imaginary parts will cancel, and we have the same result: if the roots have negative
real parts, then ai > 0.
In summary:
1. If any ai = 0, then there are roots not in the left half-plane.
2. If any ai < 0, then at least one root is in the right half-plane.
5.1
Routh-Hurwitz
The Routh-Hurwitz method uses the ideas above to determine whether a given polynomial has
roots in the right half-plane. Given the equation
an sn + an1 sn1 + + a0 = 0
27
STABILITY CRITERIA
where
b1
b2
...
c1
c2
...
an2
an5
an3
b2
an5
b3
a
n
=
an1 an1
1 an
=
an1 an4
a
n1
b1
1 an1
=
b1 b1
1
=
b1
an2
an3
array.
showing that all three coecients must be positive, which is alredy obvious from the stadnard
solution. For the cubic a3 s3 + a2 s2 + a1 s + a0 , the Routh-Hurwitz array is
a3
a1
s3
2
s
a2
a0
1
s a1 a3 a0 /a2
s0
a0
showing that the coeecients must be positive and, in addition, a1 a2 > a3 a0 , for stability.
28
STABILITY CRITERIA
Routh-Hurwitz provides a simple way of deciding whether all roots of the CE lie in
the negative half-plane.
Advantage
Routh-Hurwitz does not provide any other information about the location of the
roots: it is therefore not useful for system design.
Disadvantage
5.2
Nyquist
The Nyquist criterion is based on a polar map of the open-loop transfer function, G(j)H(j).
The Nyquist diagram is obtained by the mapping s 7 1 + G(s)H(s) where s has the following
locus:
s starts at j and moves up the imaginary axis to j. If there are poles of G(s)H(s) on
the imaginary axis, the locus follows a small semicircle, with positive real part, to go around
them.
u2 + v2 ,
The locus is innite only for generality; all that matters is that it must enclose all poles and zeros
that lie in the right half-plane.
The Nyquist criterion makes use of a theorem of Cauchy:
If F(s) is analytic within a closed contour C, then N = Z P, where N is the number of
times F(s) encircles the origin as s moves around C, Z is the number of zeroes within
C, and P is the number of poles within C.
The number of times that 1 + G(s)H(s) encircles the origin is clearly the same as the number of
times G(s)H(s) encircles the point 1 + j0. Since the contour we have chosen encloses the right
half-plane, Z is the number of zeros with positive real parts which, for stability, must be zero.
Also, P is the number of poles which, since the open-loop transfer function should be stable, is
also zero.
Definition 5.3 Nyquist Criterion As s moves along
of G(s)H(s) must not encircle the point 1 + j0.
The importance of the Nyquist criterion is that it tells us not only whether a system is stable, but
how far it is from instability. Suppose that, for some value of s, G(s)H(s) = , where is real and
> 1. Then the open-loop gain could be increased by a factor of 1 before instability occurs; the
gain margin is
1
20 log10 db
29
STABILITY CRITERIA
5.3
Bode
The Bode criterion is similar to the Nyquist criterion but uses approximations to simplify the
calculations needed. A Bode plot of a response function R(s) has two parts, the amplitude plot
showing |R(j)| and the phase plot showing arg R(j). The horizontal axis shows log (or, more
often, log10 ) and the vertical axis uses decibels for |R(j)| and degrees or radians for arg R(j).
Figure 14 summarizes the results for rst-order functions.
The log-log scales enable typical response functions to be approximated as straight lines, as follows:
Constant: K. The magnitude plot is constant at 20 log10 K and the phase plot is constant at 0
for K 0 or 180 for K < 0.
Integration: 1/j. The amplitude plot is 20 log10 which is a line with slope 20 db/decade
passing through 0 db at = 1. The phase plot is constant at /2.
Differentiation: j. The amplitude plot is 20 log10 which is a line with slope 20 db/decade
passing through 0 db at = 1. The phase plot is constant at /2.
a
. The amplitude plot is
Phase lag:
a + j
a
= 20 log q 1
20 log10
10
2
a + j
+1
a2
= 20 log10
2
+ 1.
a2
20 log10
2
+ 1 20 log10
2
a
a
which is a straight line with a slope of 20 db/decade crossing 0 db at = a. The two lines
(asymptotes) join at = a and the approximate Bode plot consists of two lines.
a
The phase is given by arg a+j
= tan1
a . For a, this is approximately zero and, for
a, it tends to /2. We approximate this as
< 0.1a : 0
0.1a
0.1a 10a :
2 10a 0.1a
> 10a : /2
To sketch the phase plot, we can draw the rst and the third components, and then draw a
straight line between their end points.
a + j
. By similar reasoning, the Bode plot is constant at 0 db for a and
a
increases at 20 db/decade for > a. The phase is approximated as
Phase lead:
< 0.1a : 0
0.1a
2 10a 0.1a
> 10a : /2
0.1a 10a :
30
STABILITY CRITERIA
Function
Amplitude
K
j
1/j
1 + j/n
20 log10 |K|
20 log10
20 log10
0
20(log10 log10 n )
0
20(log10 log10 n )
1
1+j/n
n
> n
n
> n
:
:
:
:
Point
db
1
1
n
0
0
0
Phase rule
0
< 0.1n
> 10n
< 0.1n
> 10n
:
:
:
:
/2
/2
0
/2
0
/2
1
.
G(p )
or
log G(g ) = 0.
and the phase margin is
5.4
arg G(g ) + .
Root-Locus
Root-locus techniques provide a way of studying the response of a system as one real parameter
is varied. In principle, any parameter can be varied; in practice, it is often the gain that is varied,
because this makes it easy to sketch the loci.
The root-locus is the set of paths traced by the roots of the equation 1 + KG(s) = 0 on the
complex plane with 0 K . K is the gain of the system. The equation is equivalent to the
two real equations
|KG(s)| = 1
arg KG(s) =
which are called the magnitude criterion and the phase criterion respectively. The rst
equation does not tell us much (since 0 K , it can be satised by arbitrary vlues of s).
31
STABILITY CRITERIA
The second is more useful: it says that KG(s) is a negative real number. The root locus can be
estimated by using the following rules (Shiners, pages 146{151, Philips & Harbor, pages 211.).
Figure 15 summarizes the rules for sketching the root locus: detailed notes follow.
1. There is one locus for each pole of 1 + KG(s). In other words, the number of loci is the same
as the order of the characteristic equation.
2. If the coecients of the CE are real, as is usually the case, roots are either real or complex
conjugate pairs. Consequently, each locus is symmetrical with respect to the real axis.
+
with = n m > 0. To make the number of zeros
3. In general, KG(s) has the form K bmsns+
match the number of poles, we say that KG(s) has zeros at infinity. We can construct
asymptotes to the zeros at innity, as shown below. Since the number of poles and zeros is
then equal, we have the following rule:
m
Kbm
s
= 0
d
ds G(s)
7. Let i be the angle at which a locus departs from pole pi and let j be the angle at which a
locus arrives at zero zj . Then:
j =
i +
j =
X
i
i + r
i6=j
i +
X
i6=j
i + r
32
Rule
Note
#loci = #roots of 1 + KG = 0
1
Loci are symmetrical about the real axis
2
Loci start at poles and end at zeros
3
Loci includes all real points to the left of an
4
odd number of real critical frequencies
Loci are asymptotic to zeros at innity
5
Breakaway points occur at multiple roots
6
Angles sum to 180
7
Asymptotes intersect real axis
8
Figure 15: Summary of root locus rules
where r = 1, 2, . . ..
8. The intersection of the asymptotic lines and the real axis occurs at x + j0 where
P
poles zeros
#poles #zeros
x =
The motor controller is used as a running example throughout these notes. The purpose of the
system is to control the angular position of a shaft, (t), with an electric motor. The motor is
driven by a voltage, u(t), and its equation of motion is
I
d2
dt2
d
+ ku
dt
where I is the moment of inertia of the motor, is a frictional factor, and k is the coupling constant
relating voltage to torque. The equation is in Joules (kg m2 s2 ). The units for individual variables
are:
I : kg m2
: kg m2 s1
k : Joules/volt = coulombs = A s
u : volts = kg m2 s3 A1
33
and hence
G=
k
=
radians/volt.
U
s(Is + )
Feedback is provided by a potentiometer driven by the output shaft. This provides a voltage signal
proportional to the angular position, v = f. The transfer function of the feedback loop is therefore
H = f volts/radian and the closed-loop transfer function of the system is
M(s) =
=
=
G(s)
1 + G(s)H(s)
k/s(Is + )
1 + fk/s(Is + )
k
radians/volt
2
Is + s + fk
Is2 + s + fk = 0.
fk
.
s(Is + )
(14)
Suppose the input to the system is a step function of V volts. Then U(s) = V/s and
(s) = U(s)M(s)
Vk
=
2
s(Is + s + fk)
Let
(15)
Then
(s) =
Vk
I
1
1
1
+
+
ABs A(A B)(s + A) B(B A)(s + B)
and
(t) =
Vk
I
1
eAt
eBt
+
+
.
AB A(A B) B(B A)
A =
B =
34
Since R(A) > 0 and R(B) > 0, the zeroes of the characteristic equation have negative real parts
and the system is stable. There will be oscillation if
We have
(0) =
Vk
I
<
4fk
, that is, if < 2 fkI.
I
1
1
1
+
+
AB A(A B) B(B A)
= 0
Vk
V
= .
IAB
a
lim (t) =
6.1
2
Zero Input
Consider the case when there is no input voltage. Then u = 0 and the equation of motion for the
motor shaft is
I
d
d2
+
dt2
dt
(16)
= 0.
d +
0 )) + (s(s) (0+ )) = 0.
d(
_ + ) = 0 , this simplies to
If we assume (0+ ) = 0 and (0
I0
s(Is + )
(s) =
A
B
+
.
s
Is +
This equation holds for all values of s. Setting s = 0 gives A = 1/ and setting s = /I gives
B = I/. Since I/ has the dimensions of time, we also dene = I/. Then:
(s) =
=
I0
I0
s
(s + /I)
0
0
.
s
s + 1/
The inverse Laplace transform of (s) gives the shaft position as a function of time
1
s
(t) = 0 L1
= 0 1 et/
L1
1
s + /I
35
6.2
Stability
Routh-Hurwitz
s2 I fk
s1
s0 fk.
fk
s(Is + )
fk
.
+ j
I2
The locus does not encircle the point 1 + j0 and the system is therefore stable.
6.3
Compensation
Applying compensation
s+a
gives
s+b
G =
k(s + a)
s(Is + )(s + b)
and so
GH =
fk(s + a)
s(Is + )(s + b)
s3
I
b + fk
2
s Ib +
fka
s1
B
s0
fka
36
where
1
I
b + fk
B =
fka
Ib + Ib +
=
The system is stable if B > 0 or (Ib + )(b + fk) > Ifka. Since this is equivalent to the condition
(17)
a sucient condition for stability is a < b. This is lead compensation (Section 4.2.1 on page 15)
and generally improves stability. However, we could use lag compensation provided that (17) is
satised.
6.4
Normal Form
k
1
k
can be written in the form 2
+ s + fk
I s + Is+
1
paring this to the canonical form 2
, we see that
s + 2n s + 2n
Is2
fk
I
. Com-
n =
=
fk
I
2 Ifk
1
n
2 Ifk
gives
If we put these values and b = 10 (corresponding to a time constant of 0.1 s) into (17), we obtain
the approximate condition
0.1 + 25 108 (10 a) > 0
37
SI UNITS
SI Units
Figure 16 shows the basic SI units and Figure 17 shows units derived from the basic units. We
have
RF = V A1 C V1
= A s A1
= s
showing that the product of resistance and capacitance is a time. Thus, in circuit analysis, RC is
a time and RC is a pure number.
Similarly,
HF = Wb A1 C V1
= V s A1 A s V1
= s2
showing that the product of an inductance and a capacitance is a squared time. In circuit analysis,
LC has dimension T 2 and LC2 is a pure number.
Furthermore, the unit of inductance is the product of resistance and time:
H = Wb A1
= V A1 s
= s
This means, for example, that a circuit with impedance jR1 R2 C behaves like an inductance. In
practice, we cannot construct a circuit from resistors and capacitors with pure imaginary conductance, but we could obtain an impedance of the form R + jR1 R2 C, corresponding to a resistor R
in series with an inductor L = R1 R2 C.
Quantity
Length
Mass
Time
Current
Temperature
Luminosity
Amount
Unit
metre
kilogram
second
ampere
kelvin
candela
mole
Symbol
m
kg
s
A
K
cd
mol
38
SI UNITS
Quantity
Unit
Symbol
SI
SI (basic)
kg
Frequency
Force
hertz
Hz
s1
newton
kg m s2
Friction
Torque
kg s1
newton-metre
Moment of inertia
Angular friction
s
1
1 2
1 1
kg m2 s2
2 2
kg m2
kg m2 s1
2 1
Pressure
pascal
Pa
N m2
joule
Nm
2 2
Power
watt
J s1
2 3
Charge
coulomb
As
Potential
volt
J C1
Capacitance
farad
C V1
Resistance
ohm
V A1
Conductance
siemens
A V1
Magnetic ux
weber
Wb
Vs
Flux density
tesla
Wb m2
Inductance
henry
Wb A1
1 1 2
2 3 1
2
2 3 2
2
2 2 1
2
2 2 2
39
SI UNITS
m
`
1
t
V
IR
IR
V
Rt
L
R
L
L
Rt
L
R
t2
CL
1
CL2
L
CR2
t
CR
1
CR
CR2
L
CL
t2
CL2
CR
t
CR
1
t2
t2
IL
V
Rt2
L
R
L2
L
R
L2
R2 t
L2
R2
t2
CR
1
CR2
CV
I
CL
t
CL
CR
C2 R2
t
C2 R2
V2
IR2
I
t2 2
I
t
It
It2 2
I2 R
V
Vt
L
V
L
V
R
CV
t
CV
Dimensions: T .
Dimensions: A.
IL
t
V
t2 2
I
C
IL
V
t
IR
I2 R2
V
Vt Vt2 2
40
SI UNITS
V
I
IR2
V
R2 t
L
R2
L
L
2
t
L
t
Lt2
L2
Rt2
L2 2
R
R
t
Rt
Rt2 2
t2
C2 R
C2 R2
1
Ct2
t
C
t2
C
1
C
L
CR
CR2
t
CR2
t2 2
1
41
USEFUL FORMULAS
Useful Formulas
Function
Transform
Remarks
f(t)
F(s)
General notation
Af(t) + Bg(t)
AF(s) + BG(s)
Linearity
f(t T )u(t T )
esT F(s)
eat f(t)
F(s + a)
f 0 (t)
sF(s) f(0+ )
f 00 (t)
Rt
0 f()d
Rt
0 f(t )g()d
P(s)
in the form
Q(s)
A1
A2
A3
+
+
+
s + s1 s + s2 s + s3
then
n
X
Ai
=
s + si
P(s)
Q(s)
i=1
where
Ai =
P(s)
(s + si )
.
Q(s) s=si
= +
B1
B2
Br
+
+ +
+
s + si (s + si )2
(s + si )r
where
Br =
Br1 =
P(s)
(s + si )
Q(s) s=si
1 d
r P(s)
(s + si )
1! ds
Q(s) s=si
r
42
USEFUL FORMULAS
Function
Transform
Remarks
(t)
Dirac's -function
1
s
1
sn
1
s+a
1
s(s + a)
s
2
s + a2
s
2
s a2
a
s2 + a2
a
2
s a2
1
2
s(s + a2 )
1
2
2
s (s + a2 )
1
(s + a)2
s
(s + a)2
1
2
s + 2 + 2
u(t)
tn1
(n 1)!
eat
1 eat
a
cos at
cosh at
sin at
sinh at
1 cos at
a2
at sin at
a3
teat
eat (1 at)
eat sin bt
b
a = , b = 1 2 , and < 1
p
Br2 =
1 d2
2! ds2
..
.
B1 =
r P(s)
(s + si )
Q(s) s=s
dr1
1
(r 1)! dsr1
r P(s)
(s + si )
Q(s) s=s
E(s)
R(s)
6
G(s)
B(s)
H(s)
C(s)
43
USEFUL FORMULAS
Open-loop response
Closed-loop response
G(s) H(s)
G(s)
1 + G(s) H(s)
1 + G(s) H(s) = 0
Characteristic equation
The phase-lead compensator is a form of high-pass lter. The compensator introduces gain at high
frequencies, which may increase instability, and phase lead, which tends to be stabilizing. The
pole and zero are typically placed in the high frequency region.
The phase-lag compensator is a simple form of low-pass lter. It reduces the gain at high frequencies, which tends to stabilize the system, and introduces phase lag, which tends to destabilize the
system. The pole and zero of the compensator are usually placed in the low frequency region.
Compensation
Advantages
Phase lag
LF characteristics improved
stability marrgins maintained or
improved
bandwidth reduced (useful if HF
noise is a problem)
improved stability margins
improved HF response
may be required for stability
Phase lead
Disadvantages
at least one slow term in transient reponse
reduced bandwidth may be a disadvantage
possible HF noise problems
may generate large signals, out of
linear range of system
Z2
Vout
v
Vin
Z1
Z2
Vout
Vin
Z
Z
+
Z
Z1
(b) T (s) =
2n
.
s2 + 2n s + 2n
Z1
Z1 +Z2
44
USEFUL FORMULAS
Linear
Symbol
Position
x
Mass
m
Velocity v = dx
dt
2
Force
f = m ddt2x
Spring
k df
dt = v
Friction
v = f
Rotating
Symbol
Angle
Moment of Inertia
I
Angular Velocity
= d
dt
2
Torque
T = I ddt2
Clockspring
dT
dt =
Angular friction
= T
Function
Amplitude
K
j
1/j
1 + j/n
20 log10 |K|
20 log10
20 log10
0
20(log10 log10 n )
0
20(log10 log10 n )
1
1+j/n
n
> n
n
> n
:
:
:
:
Electrical
Symbol
Charge
q
Inductance
L
Current
i = dq
dt
2
Potential
e = L ddt2q
Capacitance C de
dt = i
Resistance
Ri = e
Point
db
1
1
n
0
0
0
where
b1
b2
...
c1
c2
...
a
n
=
an1 an1
1 an
=
an1 an4
a
n1
b1
1 an1
=
b1 b1
1
=
b1
an2
an5
an3
b2
an5
b3
an2
an3
Phase rule
0
< 0.1n
> 10n
< 0.1n
> 10n
:
:
:
:
/2
/2
0
/2
0
/2