Linear Control Systems

Solution to Assignment #1
Instructor: H. Karimi

Issued: Mehr 10, 1389

Due: Mehr 18, 1389

Solution to Exercise 1.
(a) Using the superposition property of linear systems we can compute the transfer func-
tions from each input the the output individually and then obtain the overall transfer
function by adding the individual contributions.
Let’s start with the transfer function from r to c. Disregard the other two inputs,
i.e. set D(s) = 0 and N (s) = 0.
C(s) = K(s)G(s)E(s)
= K(s)G(s)(R(s) − B(s))
= K(s)G(s)R(s) − K(s)G(s)H(s)C(s)
rearranging:
K(s)G(s)
C(s) = R(s)
1 + K(s)G(s)H(s)
In a similar fashion, considering one input at the time, for disturbance we get:
G(s)
C(s) = D(s)
1 + K(s)G(s)H(s)
and for noise:
−K(s)G(s)H(s)
C(s) = N (s)
1 + K(s)G(s)H(s)
As a final result, the overall transfer function is:
1
C(s) = [G(s)D(s) + K(s)G(s)R(s) − K(s)G(s)H(s)N (s)]
1 + K(s)G(s)H(s)

(b) If there is no noise and no disturbance, and H(s) = 1, the error signal is given by:
E(s) = R(s) − B(s)
= R(s) − C(s)
KG(s)
= R(s) − R(s)
1 + KG(s)
1
= R(s)
1 + KG(s)

Linear Control Systems Solution to Assignment #1 Page 1 of 10

(c) If our input is the unit step, and if the conditions of the theorem are met, then the
steady-state error is the following:
1 1 1
ess = lim e(t) = lim sE(s) = lim s = lim
t→∞ s→0 s→0 1 + KG(s) s s→0 1 + KG(s)

Final Value Theorem doesn’t apply to E1 (s), since it has poles on the imaginary axis.
Actually, e1ss doesn’t exist, and e1 (t) oscillates sinusoidally about 0.

1 s(s + 2)
e2ss = lim K
= lim =0
s→0 1+ s(s+2)
s→0 s2 + 2s + K

Solution to Exercise 2.

(a) Input: switch-on; output: clean clothes. Open loop. The washing machine does not
feed back the cleanliness of clothes.

(b) Input: electrical signal from readout device (e.g. laser in a CD player); output: mi-
crophone vibration. Open loop. The output audio level and quality is not used to
determine the input to the speaker.

(c) Input: desired temperature; output: temperature. Closed loop. The vast majority
of air conditioners have temperature sensors (often, a simple bimetal switch) so they
turn themselves on when the temperature is outside a prescribed range, off when it
is inside the prescribed range.

(d) Manual gear train in an automobile

Input: lever location set by driver; output: wheel RPM. Open loop. The wheel speed
is not used in determining the gear setting (though a good driver would probably
apply a sense of the vehicle’s present speed to decide the proper gear to apply so as
not to stall or rev up the engine.)
Automatic gear train in an automobile
Unlike the manual gearbox, in the automatic gearbox (as the name implies!) there is
a feedback mechanism: the wheel RPM heats up a special fluid causing it to expand
or contract. The engine gear ratio is changed due to the fluid volume change.

Solution to Exercise 3.

(a) From force balance, you can derive the equation of motion. For simplicity, the system
variable θ(t) is chosen with polar coordinates. Then you don’t need to care about
tension on the rod and centrifugal force.
Free-body diagram:
Assuming that the length of the rod is l, we obtain

ml2 θ̈(t) = −mgl sin θ(t) + f (t)l cos θ(t)

Note that inertia of the mass with respect to the rotation axis is ml2 . It is a nonlinear
differential equation because it has sin θ(t) term.
If θ(t) is small enough to be assumed sin θ(t) ∼ θ(t) and cos θ(t) ∼ 1, then it is linear
equation as mlθ̈(t) = −mgθ(t) + f (t).

Linear Control Systems Solution to Assignment #1 Page 2 of 10

 Figure 1: Problem 3(a)

(b) You have one input f (t) and two outputs x(t) and θ(t), which are related to each
other. Unlike rolling with no-slipping case, you have to consider slipping because
viscous friction happens when the relative velocity at the interface of two surface is
nonzero.
Free-body diagram of the mass:

Figure 2: Problem 3(b), Free-body diagram of the mass

Free-body diagram of the inertia:

Figure 3: Problem 3(b), Free-body diagram of the inertia

The viscous friction between the mass and inertia is proportional to the relative
velocity (ẋ − rθ̇). From free-body diagrams, the equations of motion are

M ẍ(t) + 2fv ẋ(t) + Kx(t) − fv rθ̇(t) = f (t)

J θ̈(t) + r2 fv θ̇(t) − rfv ẋ(t) = 0

Linear Control Systems Solution to Assignment #1 Page 3 of 10

 If you want to obtain the relation of single input f (t) and single output x(t), then you
can eliminate x(t) and combine the above two equations. From the second equation,
J
ẋ(t) = θ̈(t) + rθ̇(t)
rfv

This implies ẍ(t) = rfJv θ(3) (t) + rθ̈(t) and x(t) = rfJv θ̇(t) + rθ(t). (Zero initial con-
dition is assumed. If not, there is only offset θ0 and it doesn’t make any difference
physically.) Using those relations, the equation of motion can be simplified to

JM θ(3) (t) + fv (r2 M + 2J)θ̈(t) + (fv 2 r2 + KJ)θ̇(t) + Kr2 fv θ(t) = fv rf (t)

It is a linear differential equation because it has only θ(t), θ̇(t), θ̈(t) and θ(3) (t) terms.

Solution to Exercise 4.

(a) Linear
If the differential equation is linear and two functions x1 (t) and x2 (t) are satisfied
with
µ ¶
2π
7ẍ1 + 0.5ẋ1 + 5 sin t x1 = f1 (t)
10
µ ¶
2π
7ẍ2 + 0.5ẋ2 + 5 sin t x2 = f2 (t)
10

then x1 + x2 also should be the solution of the differential equation with the input
f1 (t) + f2 (t), which is true because
µ ¶
2π
7(ẍ1 + ẍ2 ) + 0.5(ẋ1 + ẋ2 ) + 5 sin t (x1 + x2 ) = f1 (t) + f2 (t)
10

(b) Nonlinear
With the same method as (a), the two functions x1 (t) and x2 (t) are the solutions of

7ẍ1 + 0.5ẋ1 + 5(1 + 0.1x1 )x1 = f1 (t)

7ẍ2 + 0.5ẋ2 + 5(1 + 0.1x2 )x2 = f2 (t)

However, x1 + x2 is not the solution of the differential equation with the input f1 (t) +
f2 (t) because

7(ẍ1 + ẍ2 ) + 0.5(ẋ1 + ẋ2 ) + 5(1 + 0.1x1 )x1 + 5(1 + 0.1x2 )x2
6= 7(ẍ1 + ẍ2 ) + 0.5(ẋ1 + ẋ2 ) + 5(1 + 0.1(x1 + x2 ))(x1 + x2 )

(c) Nonlinear

µ ¶
d 1 2 1 2
mẋ + kx = 0
dt 2 2
⇒ ẋ(mẍ + kx) = 0
⇒ ẋ = 0 and/or mẍ + kx = 0

Linear Control Systems Solution to Assignment #1 Page 4 of 10

 Let A and B denote the solution space of the homogeneous linear differential equations
ẋ = 0 and mẍ + kx = 0, respectively. Clearly, A4B 6= ∅. Choose x ¡ 1 (t) ∈ A − A4B¢
d 1 2 1
and x2 (t) ∈ B − A4B; then both x1 (t) and x2 (t) are solutions of dt 2
m ẋ + 2
kx2 =
0, but xs (t) = x1 (t) + x2 (t) is not, since it satisfies none of the differential³qequations
´
k
ẋ = 0 and mẍ + kx = 0. A simple example is x1 (t) = 1 and x2 (t) = sin m
t .
Some people may believe that the correct answer is “Linear” [under an assumption].1
They may justify their answer either way:

1. By taking derivative first and inspecting.

µ ¶
d 1 2 1 2
mẋ + kx = ẋ(mẍ + kx) = 0
dt 2 2

If we assume ẋ 6= 0, then the equation is linear. (The reason why the assumption
is valid, if ẋ = 0 then x = cte and trivial.)
2. With the same method, we have two functions x1 and x2 satisfied with
µ ¶
d 1 2 1 2
mẋ1 + kx1 = 0
dt 2 2
µ ¶
d 1 2 1 2
mẋ2 + kx2 = 0
dt 2 2

From the first equation, and again assuming ẋ1 6= 0, ẋ2 6= 0 for nontrivial solution,

mẍ1 + kx1 = 0

From the second equation

mẍ2 + kx2 = 0

Now consider
· ¸
d 1 2 1 2
m(ẋ1 + ẋ2 ) + k(x1 + x2 )
dt 2 2
· ¸
d 1 2 2 1 2 2
= m(ẋ1 + ẋ2 + 2ẋ1 ẋ2 ) + k(x1 + x2 + 2x1 x2 )
dt 2 2
= m(ẋ1 ẍ1 + ẋ2 ẍ2 + ẋ1 ẍ2 + ẍ1 ẋ2 ) + k(x1 ẋ1 + x2 ẋ2 + x1 ẋ2 + ẋ1 x2 )
= ẋ1 (mẍ1 + kx1 ) + ẋ2 (mẍ2 + kx2 ) + ẋ1 (mẍ2 + kx2 ) + ẋ2 (mẍ1 + kx1 )
=0

Since the input x1 + x2 also satisfies the equation of motion, the system is linear.

Solution to Exercise 5. Let’s define φ(t) as the rotation angle of the motor. From the
relation of the gear pair, we know φN1 = θN2 and φ̇N1 = θ̇N2 . T is the torque generated
by the motor and it is scaled by N2 /N1 at the inertia by the gear pair. Also the inertia
of the motor is scaled by N22 /N12 by the gear pair. (For the detail, please refer to section
2.7 of Nise)
1
In my opinion, their assumption is not valid.

Linear Control Systems Solution to Assignment #1 Page 5 of 10

From KCL at the node attached to the resistor,
ve T
is (t) − − = 0.
R Km
From torque balance at the inertia,
µ ¶
N22 N2
J + 2 Jm θ̈ = T − bθ̇ − Kθ.
N1 N1

(Converted torque from the motor drives the effective inertia (inertia + converted inertia
of the motor shaft), viscous friction and compliance.)
N2
Using the Laplace transform and ve = Kv φ̇(t) = Kv N 1
θ̇(t), the above two equations can
be written as
K v N2 T (s)
Is (s) −
Θ(s)s − =0
R N1 Km
µµ ¶ ¶
N22 2 N2
J + 2 Jm s + bs + K Θ(s) = T (s)
N1 N1
Solving the first equation with respect to T and plugging it into the second equation, we
get the transfer function defined by
Θ(s) (N2 /N1 )Km
=³ ´ ³ ´ .
Is (s) J+
N22
J 2 N22 Kv Km
s + b + N2 R s+K
N12 m 1

Solution to Exercise 6.
dF (s)
(a) Use L{tf (t)} = − twice:
ds
(s + 1)3
F (s) = .
6

(b) Note that sin t cos t = 12 sin(2t).2 Then,

1
F (s) = .
s2 + 4

dF (s)
(c) Use the identity sin(t − 3) = sin t cos 3 − cos t sin 3 and L{tf (t)} = − :
ds
cos 3 2s sin 3 (s2 + 1) − sin 3 2s2
F (s) = + .
(s2 + 1)2 (s2 + 1)2

dF (s)
(d) Here you should use L{tf (t)} = − first and then ∀T > 0, L{f (t−T )1(t−T )} =
ds
e−T s F (s):

e−3s
F (s) = .
(s − 1)2
2
avval ya dovome dabirestan!

Linear Control Systems Solution to Assignment #1 Page 6 of 10

(e) Since
1
L{te−at } =
(s + a)2
s2 − a2
L{t cos at} =
(s2 + a2 )2
hence
L{te−at + 2t cos t} = L{te−at } + L{2t cos t}
1 s2 − 1
= + 2
(s + a)2 (s2 + 1)2

Solution to Exercise 7.
s2 + s + 1
(a) Let F1 (s) = . Write
(s + 1)(s + 2)(s + 3)
a1 a2 a3
F1 (s) = + + ;
s+1 s+2 ¯ s+3
s2 + s + 1 ¯¯ 1
a1 = ¯ = ,
(s + 2)(s + 3) s=−1 2
¯
s2 + s + 1 ¯¯
a2 = = −3,
(s + 1)(s + 3) ¯s=−2
¯
s2 + s + 1 ¯¯ 7
a3 = ¯ = .
(s + 1)(s + 2) s=−3 2
µ ¶
1 −t −2t 7 −3t
⇒ f1 (t) = e − 3e + e 1(t).
2 2
Since ∀T > 0, L{x(t − T )1(t − T )} = e−T s X(s), hence
f (t) = f1 (t − 1)1(t − 1)
µ ¶
1 −(t−1) −2(t−1) 7 −3(t−1)
= e − 3e + e 1(t − 1).
2 2

(b)
½ ¾ ½1 1 1 ¾
−1 1 −1 4 4 2
L =L − −
s(s + 2)2 s (s + 2)2
s+2
1© ª
= 1 − e−2t − 2te−2t 1(t)
4
(c) It can be re-written as
µ ¶
2 2
F (s) = s − 3s + 1 + H(s).
s
Therefore its inverse Laplace transform is
Z t
d2 h(t) dh(t)
f (t) = −3 + h(t) + 2 h(τ ) dτ.
dt dt 0−

Linear Control Systems Solution to Assignment #1 Page 7 of 10

(d)

ωn2
F (s) =
s (s2 + 2ζωn s + ωn2 )
1
= F1 (s)
s
ωn2
F1 (s) = 2
s + 2ζωn s + ωn2

For all 0 < ζ < 1,

f1 (t) = L−1 {F1 (s)}

ωn ³ p ´
=p e−ζωn t sin ωn 1 − ζ 2 t
1 − ζ2
1 ³ p ´
⇒ f (t) = 1 − p e−ζωn t sin ωn 1 − ζ 2 t + cos−1 ζ
1 − ζ2

Hint: You should use “integration by parts” twice.

(e)

1 − e−s
F (s) =
s2 (1 + es )
1
= 2 F1 (s)
s
1 − e−s
F1 (s) =
1 + es
e−s (1 − e−s )
=
e¡−s + 1 ¢ ¡ ¢
= e−s 1 − e−s 1 − e−s + e−2s − · · ·
= e−s − 2e−2s + 2e−3s − · · ·
⇒ f1 (t) = δ(t − 1) − 2δ(t − 2) + 2δ(t − 3) − · · ·
RR
=⇒ f (t) = (t − 1)1(t − 1) − 2(t − 2)1(t − 2) + 2(t − 3)1(t − 3) − · · ·

Note that expanding the term 1/(1 + es ) [directly] to a geometric series, leads to an
invalid non-causal answer; we are using one-sided Laplace transform definition.

Solution to Exercise 8.

Linear Control Systems Solution to Assignment #1 Page 8 of 10

(a)
µ ¶µ ¶
1 1
Y (s) =
s+1 s2 + 2s + 5
· ¸· ¸
1 1
=
s + 1 (s + 1 + 2j)(s + 1 − 2j)
c1 c2 c3
= + +
s + 1 s + 1 + 2j) s + 1 − 2j)
1 1 1
c1 = . =
2j −2j 4
1 1 1
c1 = . =−
−2j −4j 8
1
c1 = c∗2 = −
· 8 ¸
1 −t 1 (−1−2j)t 1 (−1+2j)t
y(t) = e − e − e 1(t)
4 8 8
· ¸
1 −t 1 −t ¡ −2jt ¢
= e − e e + e2jt 1(t)
4 2
1 −t
= e [1 − cos 2t] 1(t)
4

(b)
3s + 4 1
Y (s) =
s4
+ 4s3
2
+ 8s + 8s + 4 s
1 3s + 4
=
s [(s + 1)2 + 1]2
k c1 c2 d1 d2
= + + + +
s s + 1 + j (s + 1 + j)2 s + 1 − j (s + 1 − j)2

k=1
1 1 1 1
c1 = − − j c2 = − − j
2 4 4 2
1 1 1 1
d1 = c∗1 = − + j d2 = c∗2 = − + j
2 4 4 2
µ· ¶ µ ¶
1 1 (−1−j)t 1 1
y(t) = 1 + − − j e + − − j te(−1−j)t
2 4 4 2
µ ¶ µ ¶ ¸
1 1 (−1+j)t 1 1 (−1+j)t
+ − + j e + − + j te 1(t)
2 4 4 2
· µ ¶¸
−t 1 1
= 1−e cos t + sin t + t cos t + t sin t 1(t)
2 2

Solution to Exercise 9 (Nise, 4th Edition, Chapter 2, Problem 53). The relationship
between the nonlinear spring’s displacement, xs (t), and its force, fs (t), is

xs (t) = 1 − e−fs (t) .

Linear Control Systems Solution to Assignment #1 Page 9 of 10

Solving for the force,

fs (t) = − ln(1 − xs (t)) (1)

Writing the differential equation for the system by summing forces,

d2 x(t) dx(t)
+ − ln(1 − x(t)) = f (t). (2)
dt2 dt
Letting x(t) = x0 + δx and f (t) = 1 + δf , linearize ln(1 − x(t)):
¯
d ln(1 − x) ¯¯
ln(1 − x) − ln(1 − x0 ) = ¯ δx.
dx x=x0

Solving for ln(1 − x),

¯
d ln(1 − x) ¯¯ 1
ln(1 − x) = ln(1 − x0 ) − ¯ δx = ln(1 − x0 ) − δx. (3)
dx x=x0 1 − x0

When f = 1, δx = 0. Thus from (1), 1 = − ln(1 − x0 ). Solving for x0 ,

1 − x0 = e−1
x0 = 0.6321.

Substituting x0 = 0.6321 into (3),

1
ln(1 − x) = ln(1 − 0.6321) − δx = −1 − eδx
1 − 0.6321
Placing this value into (2) along with x(t) = x0 + δx and f (t) = 1 + δf , yields the
linearized differential equation

d2 δx dδx
+ + 1 + eδx = 1 + δf
dt2 dt
or
d2 δx dδx
+ + eδx = δf.
dt2 dt
Taking the Laplace transform and rearranging yield the transfer function

∆X(s) 1
= 2
∆F (s) s +s+e

Good Luck!

Linear Control Systems Solution to Assignment #1 Page 10 of 10