Professional Documents
Culture Documents
Effects PDF
Effects PDF
systems
Example Suppose the plant is quite slow, its time constant is τ orig = 100 sec,
1
ẋ = −0.01x + u, or, equivalently, X(s) = U (s).
s + 0.01
To speed up it to τdesired = 1 sec, use the feedback controller
u = −kx + r.
Then, the closed loop system is
1
ẋ = −(0.01 + k)x + r, or, equivalently, X(s) = R(s).
s + 0.01 + k
To achieve the desired time constant of 1 sec, the pole of the closed loop system must be placed
at s = −1/τdesired = −1, i.e., we need a = (0.01 + k) = 1, which is achieved using the gain of
k = 0.99.
1
R U Y
+
Plant
−
K
PSfrag replacements
Feedback Controller
Note on the form of the system. Eq. (1) is selected because its value at s = 0 equals 1 for
any p. We will see that the step response of the system will approach the value of 1 for large
values of t, independent of the value of p. This is convenient, since we wish to examine the step
response for different values of p. However, the conclusions drawn from the example will apply
to any second order system with two real poles.
Let R(s) = 1/s and perform a partial-fraction expansion on the resulting step response
transform Y (s):
p 1
p 1
Y (s) = = − p−1 + p−1 .
s(s + 1)(s + p) s s+1 s+p
The step response y(t) is given by
p −t 1 −pt
y(t) = 1 − e + e . (2)
p−1 p−1
p
We will consider this response as the sum of two terms. The first term is given by 1 − p−1
e−t ,
and the second term is p−1
1
e−pt .
Let us examine the step response of Eq. (2) for different values of p:
2
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
p=infinity
p=10
0.2 p=1
p=0.5
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure 2: Step response of the system with the additional pole at −p.
pv=[.5 1 10 inf];
t=[0:.1:5];
yv=[];
for i=4:-1:1
p=pv(i);
G=tf(1,conv([1 1],[1/p 1]));
[y,x]=step(G,t);
yv=[yv y];
end
plot(t,yv(:,1),’-’,t,yv(:,2),’--’,t,yv(:,3),’-.’,t,yv(:,4),’.’)
legend(’p=infinity’,’p=10’,’p=1’,’p=0.5’);
grid
This code produces the step response of Eq. (2) for p = 0.5, 1, 10 and ∞; see Figure 2. For
p = ∞, the step response is the same as the step response of the system with the first-order
transfer function s+1
1
. As p reduces, the plot moves away from the step response of this first-
order system.
Indeed p/(p − 1)e−t → e−t as p → ∞, and 1/(p − 1)e−pt → 0 as p → ∞. In the latter limit,
two factors matter: 1/(p − 1) p/(p − 1), and also e−pt has a time constant 1/p decreases as
p → ∞, hence this is a fast term. The conclusion can be drawn from Figure 2 that if p 1, the
term p−1
1
e−pt corresponds to the fast part of the step response. The effect of this term (and hance
the effect of the additional pole at s = −p is negligibly small as t → ∞. Furthermore, the term
p
1 − p−1 e−t describes the dominant part of the steop response. Since the response is similar to
the response of the first-order system, the system with p 1 is said to be dominantly first-order.
Also, we say that the pole at s = −1 dominates on the pole at s = −p; see Figure 3
3
Im
Re
-p -1
Dominant pole
Additional zero Im
Re
-p -z -1
Dominant pole
Figure 4: Zero-pole locus of the system with the additional zero at s = −z.
4
4.5
z=∞
4 z=10
z=2.
z=1.
3.5
z=.5
z=.2
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure 5: Step response of the dominantly first-order system with the additional zero at s = −z.
10 −t
y(t) ≈ yd (t) = 1 − e for sufficiently large t.
9
Thus, the pole at s = −1 remains dominant.
z = 10. This zero cancels the pole at s = −10. The system becomes a first-order system.
1 < z < 10. In this case, the additional zero speeds up the system; see Figure 5.
z < 1. The additional zero becomes dominant. It speeds up the system and at the same time
leads to occuring an overshoot; see Figure 5.
From this analysis, one can see the general effect that the speed of the response increases as zero
moves from +∞ to 0 along the negative real axis. When zero becomes dominant, an overshoot
occurs.
To obtain Figure 5, the following Matlab code was used:
zv=[.2 .5 1. 2. 10 inf];
t=[0:.1:5];
yv=[];
for i=6:-1:1
z=zv(i);
G=tf([1/z 1],conv([1 1],[1/10 1]));
5
[y,x]=step(G,t);
yv=[yv y];
end
plot(t,yv(:,1),’--’,t,yv(:,2),’-’,t,yv(:,3),’-.’,t,yv(:,4),’-o’,t,yv(:,5),
’-x’,t,yv(:,6),’-*’)
legend(’z=\infty’,’z=10’,’z=2.’,’z=1.’,’z=.5’,’z=.2’);
grid;
1.4 Summary
1. We have seen that if there is a single pole that is significantly closer to the origin than
other poles and zeros of a transfer function with all its poles in the left halfplane, the time
constant of that pole closest to the origin dominates the response of the system.
6
1
0.5
−0.5
−1
−1.5
−2
z=−∞
−2.5 z=−2.
z=−0.5
z=−0.2
−3
−3.5
−4
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
2. The response from the dominant pole is modified from a pure first-order system response
by the presence of other poles and zeros. Additional poles delay the response of the system
while left half-plane zeros speed up the response. Right halfplane zeros cause the response
to start off in the wrong direction before recovering. The effect increases as either a pole
or zero moves toward the origin.
3. If there is a zero and a pole modifying the effect of another dominant pole and the modi-
fying zero is closer to the origin than the modifying pole, the response from the dominant
pole is modified more by the zero than the pole and the response is slightly advanced or
sped up in time. If the modifying pole is closer to the origin than the modifying zero, the
response is modified more by the pole than the zero and the response is slightly delayed.
7
2 Effects of additional poles and zeros on performance of a
second order underdamped system
It often happens that the poles closest to the origin are a complex pole pair. Hence, the dominant
response will be from those poles which correspond to a second-order system. We will see in
this section that additional poles and zeros modify the response of a second-order system in a
similar way to how the response of a first-order system is modified.
The discussion is based upon the second-order differential equation
d2 y dy
2
+ 2ζωn + ωn2 y(t) = ωn2 r(t), (4)
dt dt
where r(t) is a step-function input and the initial conditions are assumed to be zero. The constant
ζ is called the damping ratio, and ωn is referred to as the undamped natural frequency.
There are numerous examples in which the above differential equation occurs such as a series
electrical circuit containing resistors, inductors and capacitors, a system consisting of a spring,
mass, and viscous friction etc.
2.1 The derivation of the unit step response of the prototype second order
system with two complex conjugate poles
The transfer function of the system of Eq. (4) is
Y (s) ωn2
= 2 (5)
R(s) s + 2ζωn s + ωn2
The transfer function (5) has poles located at
q
s = −ζωn ± ωn (ζ 2 − 1) (6)
R1 + R2 e−ζωn + R3 te−ζωn ;
R1 , R2 , R3 are constants.
If 0 < ζ < 1, poles are complex conjugate and located in the left half-plane:
q
s = −ζωn ± jωn (1 − ζ 2 ) (7)
We focus on the last case. The location of the poles with respect to the damping ratio and the
undamped natural frequency is indicated in Fig. 7. To investigate the response of the system, we
8
Pole jω
s−plane
ωn √
ωd = ω n 1 − ζ 2
θ
0
Re
ωn ζ
PSfrag replacements
Pole
chose the input as a step input. For R(s) = 1/s, Y(s) is given by
ωn2
Y (s) = .
s [(s + ζωn )2 + ωn2 (1 − ζ 2 )]
The partial fraction expansion is found for the case of the pair of complex conjugate poles and a
single real pole:
R1 R R̄
Y (s) = + √ 2
+ √
s s + ζωn + jωn 1 − ζ s + ζωn − jωn 1 − ζ 2
We use the residues method:
R1 = 1
ωn2
R = √ 2
√
s(s + ζωn − jωn 1 − ζ ) s=−ζωn −jωn 1−ζ 2
−j
= √ √
2(ζ + j 1 − ζ 2 ) 1 − ζ 2
!
1 ζ
= − 1 + j√
2 1 − ζ2
1 √
jtan−1 (ζ/ 1−ζ 2 )
= − √ e
2 1 − ζ2
1
= − √ ej(π/2−θ)
2 1 − ζ2
9
where
θ = cos−1 ζ.
We now use the table of Laplace transforms and write y(t) as
1 −ζωn t
q π
y(t) = 1 − √ 2
e cos(ω n t 1 − ζ2 + θ − )
1−ζ 2
1 q
= 1− √ 2
e−ζωn t sin(ωn t 1 − ζ 2 + θ)
1−ζ
√
The value ωd = ωn 1 − ζ 2 is the actual frequency of oscillation in radians per second, it is
referred to as the damped frequency. The period of oscillation, T d , associated with the damped
frequency ωd is
2π 2π
Td = = √ .
ωd ωn 1 − ζ 2
Typical responce of the system is shown in Fig. 8. In addition to the period and frequency of
Step Response
1.5
Maximum overshoot
Td
105%
1
95%
90%
Amplitude
0.5
Settling time
10%
0
0 5 10 15 20 25
Time (sec)
Rise time
Figure 8: Typical unit step response of an underdamped (0 < ζ < 1) second-order system.
damped oscillations, other important transient performance characteristics of a stable 2nd order
underdamped system are
10
− √ πζ
• Percent maximum overshoot, P O = e 1−ζ 2 × 100%;
• Rise time shows how long it takes for the response to rise from from 10% of the final value
to 90% of the final value. There is no exact equation to express the rise time.
• Settling time shows how long it takes for transients to settle. The settling time is measured
as a time required for the response to settle to within ±5%, or in some cases ±2% of the
final value. In the first case, the settling time is approximately equal to t setlle = ζω3n .
Effect of varying the damping ratio ζ Let ωn = 1 be fixed. Step responces for various
damping ratios are given in the Figure 9. The overshoot decreases as ζ increases. However
the decrease in the overshoot is at the expense of rise time. While the response becomes more
sluggish, its settles quicker as ζ increases.
Effect of varying the undamped natural frequency ωn Varying the undamped natural fre-
quency ωn simply scales the time axis, see the Figure 10. It affects the period of damped oscilla-
tions, rise and settling time, but has no effect on the percent maximum overshoot.
2.2 The effect of additional poles and zeros on a dominantly second order
system
The effects that added poles and zeros have on dominantly second-order systems are similar to
the effects that added poles and zeros have on dominantly first-order systems.
This system has the same poles as the system in equation (5). Also, it has a single zero at s = −z.
The effect of additional left-halfplane zero can be seen from the following analysis. The unit
step response of the above transfer function can be written as follows
ωn2
z
(s
+ z)
Y (s) =
s(s2 + 2ζωn s + ωn2 )
ωn 2
ωn2 z
s
= +
s(s + 2ζωn s + ωn ) s(s + 2ζωn s + ωn2 )
2 2 2
11
Pole−Zero Map
1.2
0.74 0.6 0.42 0.2
0.83
ζ=0.1
1 ζ=0.3
ζ=0.5
0.8
0.91 ζ=0.707
0.6
Imag Axis
0.96
0.4
0.99
0.2
−0.2
−1.6 −1.4 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0
Real Axis
1.8
ζ=0.1
1.6
ζ=.3
ζ=.5
ζ=.707
1.4
1.2
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25
Step responses
Figure 9: Second order system with various damping ratios. Locations of poles on the complex
plane are obtained using Matlab function pzmap. The (ζ, ω n )-grid can be added using grid or
sgrid.
12
Pole−Zero Map
1.5
0.74 0.6 0.42 0.22
ω =2
0.84 n
1 System: sys
Pole: −0.707 + 0.707i
0.91 Damping: 0.707
Overshoot (%): 4.32
Imag Axis
Frequency (rad/sec): 1
ω =1
n
0.96
0.5
ωn=0.5
0.99
0.8
0.6
ωn=0.5
ωn=1.
ωn=2.
0.4
0.2
0
0 5 10 15 20 25
Figure 10: Second order system with various undamped natural frequencies.
13
zv=[0.5 1 2 10 inf];
t=[0:.1:5];
omega_n=1;
zeta=sqrt(2)/2;
yv=[];
for i=5:-1:1
z=zv(i);
G=tf([omega_nˆ2/z omega_nˆ2],[1 2*zeta*omega_n omega_nˆ2]);
[y,x]=step(G,t);
yv=[yv,y];
end
plot(t,yv(:,1),’-’,t,yv(:,2),’--’,t,yv(:,3),’-.’,t,yv(:,4),’-o’,t,yv(:,5),’-
grid;
legend(’z=\infty’,’z=10’,’z=2’,’z=1’,’z=0.5’);
0.5
z=∞
z=10
z=2
z=1
z=0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure 11: Effect of an additional left halfplane zero on a dominantly second-order system
14
in the left half-plane makes the system faster and more oscillatory. This can be seen from the
simulations.
As the zero moves along the negative real axis toward the origin, the time to the first peak of
the step response decreases monotonically while the percent overshoot increases monotonically.
Also, it takes longer for the system to settle to the final value of the response.
The zero in the right half-plane retards the system and produces an undershoot. The persent
undershoot decreases as the zero moves along the positive real axis toward the infinity, see Fig-
ure 12. Again the system oscillates for a longer time.
0.8
0.6
0.4
0.2
0
z=−∞
z=−10
−0.2 z=−2
z=−1
z=−0.5
−0.4
−0.6
−0.8
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure 12: Effect of an additional zero in the right half-plane on a dominantly second-order
system
15
2.2.2 An additional zero in the forward path transfer function
Consider an example. Let the forward path transfer function be
Y (s) 6(zs + 1)
=
R(s) s(s + 1)(s + 10)
The effect of the pole at s = −10 is insignificant as transients due to this pole should die out
fairly quickly. The closed loop transfer function is
Y (s) 6(zs + 1)
= 3
R(s) s + 11s2 + (10 + 6z)s + 6
AS we can see, this time z affects not only the numerator, but the denominator also contains z. In
this case, a larger z (and a zero closer to the imaginary axis) has an effect of improving damping
and reducing maximum overshoot.
16
pv=[0.1 .5 2 10 inf];
t=[0:.3:15];
omega_n=1;
zeta=sqrt(2)/2;
yv=[];
for i=5:-1:1
p=pv(i);
G=tf(omega_nˆ2,conv([1/p 1],[1 2*zeta*omega_n omega_nˆ2]));
[y,x]=step(G,t);
yv=[yv,y];
end
plot(t,yv(:,1),’-’,t,yv(:,2),’--’,t,yv(:,3),’-.’,t,yv(:,4),’-o’,t,yv(:,5),’-
grid;
legend(’p=\infty’,’p=10’,’p=2’,’p=0.5’,’p=0.1’);
1.4
p=∞
p=10
1.2 p=2
p=0.5
p=0.1
0.8
0.6
0.4
0.2
0
0 5 10 15
17
Step Response
2.5
p=10
p=1
2
1.5
Amplitude
1
0.5
−0.5
p=0 (no added pole)
−1
0 5 10 15
Time (sec)
2.3 Summary
As was in the case of a dominantly first-order system, the response of a dominantly second-
order system is sped up by an additional zero and is slowed down by an additional pole. In
the dominantly second-order system the added closed loop zero also has the important effect of
increasing the amount of oscillation in the system while an added closed loop pole has the effect
of decreasing the amount of oscillation. Added forward path zeros and added forward path poles
have an opposite effect on the overshoot. A forward path pole which is too close to the origin
may turn the closed loop system unstable.
A right half-plane zero also causes a ‘wrong way’ response. All effects become more pro-
nounced as the additional zero or pole approach the origin and become dominant.
18