ELEN90055 Control Systems

The University of Melbourne

Problem Set 5 with Reference Solutions
Prepared by M. Cantoni, May 2012

x=0
x

Figure 1: Hovering mass with lateral force actuator and ultra-sound position sensor.

Problem 1
Figure 1 illustrates a mass m that hovers on a bed of air. In the plane, it is constrained to move
in the x direction only. The mass has a bi-directional propeller which enables a control system to
adjust the net force f acting on the mass in the x direction. An ultra-sound sensor is used to measure
the position relative to a datum point designated to be x = 0. Applying Netwon’s second law in
the x direction yields the following differential equation model: ẍ(t) = f (t)/m. The position of the
mass can be regulated via feedback control as shown in Figure 2, where C(s) is the transfer function
of a compensator, R(s) is the Laplace transform of a position reference signal, D(s) is the Laplace
transform of a signal that represents the difference between the commanded and actual net force,
as a simple model for the uncertain actuator characteristics, and X(s) is the Laplace transform of
the measured position signal. Suppose the performance specifications require an open-loop transfer
function that has a phase margin of 45◦ and cross-over frequency of 3 rad/sec. Let m = 1 kg.
1. Is it possible for a constant gain compensator C(s) = K to achieve closed-loop stability and
the performance specifications given above?
2. Is it possible for a lag compensator C(s) = K(τz s + 1)/(τp s + 1) with 0 < τz < τp to achieve
closed-loop stability and the performance specifications given above?
3. Is it possible to use a two-parameter lead compensator C(s) = K(τz s + 1)/((τz /9)s + 1) with
K > 0 and τz > 0 to achieve closed-loop stability and the performance specifications given
above?
4. Design a three-parameter lead compensator C(s) = K(τz s + 1)/(τp s + 1) with K > 0 and
0 < τp < τz to achieve closed-loop stability and the performance specifications given above,
in addition to the requirement that the one real closed-loop pole have magnitude 10 times the
magnitude of the real part of the two under-damped closed-loop poles.

1
 D(s)

R(s) U (s) + F (s) X(s)

+ + 1
C(s)
_ ms2

Figure 2: Position control system

Problem 2

+
F C(s) Vo
-

+ + Vf - + +
Vi Vm A(s) Vo
- - -

Figure 3: Feedback amplifier – two-port network model.

Figure 3 shows a two-port network model for an electronic feeback amplifier. The active two-port
component is labelled A(s), which denotes the incremental voltage-to-voltage (V-V) transfer function,
given by
K
A(s) = ,
(τ2 s + 1)(τ1 s + 1)
p
where τ1 = 0.1, τ2 = 0.001 and K = 2(1 + τ12 /τ22 ), so that |A(j τ12 )| = 1; the Bode plot of A(s) is
shown in Figure 4. The two-port devices labelled F and C(s) are passive. The former is a precision
attenuator (i.e. resistor divider) with V-V transfer function that is a constant F ∈ (0, 1] (Ohms) and
the latter is one of the following: (i) a direct connection Cd (s) = 1; (ii) a resistor-capacitor lead
network with transfer function
τp (τz s + 1)
C+ (s) =
τz (τp s + 1)
with 0 < τp < τz ; or (iii) a resistor-capacitor lag network with transfer function
(τz s + 1)
C− (s) =
(τp s + 1)
with 0 < τz < τp . The notation Vx (s) is used for the Laplace transform of the port voltage signal
vx (t) for x ∈ {i, o, m, f }.
Kirchoff’s voltage law gives Vi (s) = Vf (s) + Vm (s). Using this and the relationships
Vf (s) = F C(s)Vo (s) and Vo (s) = A(s)Vm (s),
it follows that the transfer function Vo (s)/Vi (s) = A(s)/(1+F C(s)A(s)). The open-loop transfer func-
tion is Λ(s) = F C(s)A(s). Note that the steady-state gain for a step input is A(0)/(1+F C(0)A(0)) ≈
1/F C(0) since A(0) is large. When C(s) is Cd (s) or C− (s) then C(0) = 1 and the over all amplifier
τ
gain is approximately 1/F . When C(s) = C+ (s) then C(0) = τpz < 1, whereby the overall gain at low
frequency is greater than 1/F .

2
 Bode Diagram
45
40
35
30
25 A
20
Magnitude (dB)

15
10
5
0
−5 −C+
−10
−15
−20 C−
−25
−30
0
−10 C−
−20
−30
−40
−50
−60
Phase (deg)

−70
−80 A
−90
−100
−110
−120
−130
−140
−150 −C+
−160
−170
−180 0 1 2 3 4
10 10 10 10 10
Frequency (rad/s)

Figure 4: Bode plots of A(s), −C+ (s) and C− (s)

3
 1. Suppose that C(s) = Cd (s) = 1. Using the Bode plots in Figure 4, rather than direct calculation,
find the values of the constant F for which the phase margin is greater than 90◦ ?

2. Suppose that C(s) = C− (s) with the parameters τp , τz set in line with the corresponding Bode
plot shown in Figure 4. Using the Bode plots, rather than direct calculation, find the values of
the constant F for which the phase margin is greater than 90◦ ?

3. Suppose that C(s) = C+ (s) with the parameters τp , τz set in line with the corresponding Bode
plot (of −1 × C+ (s)) shown in Figure 4. Using the Bode plots, rather than direct calculation,
find the values of the constant F for which the phase margin is greater than 90◦ ?

Reference Solutions:

Problem 1
1. It is not possible to achieve closed-loop stability using a constant gain compensator C(s) = K.
To see this, consider the closed-loop characteristic
√ polynomial, which would be s2 + K in
this case. The roots of this are s = ±j K, which are purely complex. That is, a constant
gain compensator can only achieve a marginally stable closed-loop and so the performance
specifications do not make sense as these presume closed-loop stability.

2. A lag compensator C(s) = K(τz s + 1)/(τp s + 1) with 0 < τz < τp cannot achieve closed-loop
stability. To see this, consider a root-locus plot. The open-loop poles are at s = 0, 0, −1/τp and
the open-loop zero is at s = −1/τz . As the gain K increases from 0, the closed-loop pole that
originates at the open-loop pole s = −1/τp , moves further into the left-half plane towards the
open-loop zero at s = −1/τz . The relative degree of the open-loop transfer function C(s)/s2
is 2. So the two closed-loop poles that originate at the two open-loop poles s = 0, each move
towards infinity along asymptotes with angles of ±90◦ for large K; these asymptotes intersect
the real axis at (−1/τp + 0 + 0 − (−1/τz ))/2 > 0, in the right-half plane. The angles of departure
of the two closed-loop poles from the origin as K increases from 0 are given by ±90◦ , since near
the origin ∠(s + 1/τz ) ≈ ∠(s + 1/τp ) ≈ 0◦ and if the “phase condition”

∠(s + 1/τz ) − ∠s2 − ∠(s + 1/τp ) = (1 + 2ℓ)180◦ (ℓ = 0, ±1, . . .)

holds, it follows that ∠s2 = 2∠s = ±180◦ . The two closed-loop poles then move into the
right-half plane, towards the asymptotes. This is unstable.

3. A two-parameter lead compensator C(s) = K(τz s + 1)/((τz /9)s + 1) with K > 0 and τz > 0
can be used to achieve closed-loop stability. To see this, consider a root-locus plot. As the gain
K increases from 0, the closed-loop pole that originates at the open-loop pole s = −9/τz moves
closer to the imaginary axis, towards the open-loop zero at s = −1/τz . Again, the two closed-
loop poles that originate at s = 0 each move towards infinity along asymptotes with angles of
±90◦ for large K; these asymptotes intersect the real axis at (−9/τz + 0 + 0 − (−1/τz ))/2 < 0,
in the left-half plane. The angles of departure are again ±90◦ , but the closed-loop poles then
move into the left-half plane towards the asymptotes, which is stable. Therefore, in this case,
it make sense to check if the performance specification can be met.
The phase margin and cross-over frequency specifications correspond to the following constraints
on the open-loop transfer function:
s
K (jτz 3 + 1) K 9τz2 + 1
1= =
(j3)2 (jτz /3 + 1) 9 τz2 /9 + 1

4
 and
 
1 (jτz 3 + 1)
−135 = −180 + 45 = ∠
◦ ◦ ◦
(j3)2 (jτz /3 + 1)
 
(jτz 3 + 1)(−jτz /3 + 1) 8τz
= −180 + ∠
◦
2
= −180◦ + arctan .
τz /9 + 1 3(1 + τz2 )

This last conditionpimplies 8τz /(3 + 3τz2 ) = tan(45◦ ) = 1 or equivalently 3τz2 − 8τz + 3 = 0.
That is, τz = (8 ± (64 − 36))/6. The gain condition can then be used to find a corresponding
K for these two possible choices for τz > 0; a different value of K is required for each choice.
One choice places the cross-over frequency before the maximum phase lead provided. The other
choice places the cross-over frequency after the maximum phase lead provided.

4. For a three-parameter lead compensator C(s) = K(τz s + 1)/(τp s + 1) closed-loop stability

can be achieved for the same reasons described above in the case of a two-parameter lead
compensator. The open-loop transfer function C(s)/s2 has three poles and so the closed-loop has
three poles; the characteristic polynomial is degree 3. By the fundamental theorem of algebra
one of these is real valued and the other two are either both real or a complex conjugate pair.
The additional performance specification on the one real pole ensures dominance of the other
two as a complex conjugate pair, which should be set to have damping and natural frequency
characteristics consistent with the phase margin and cross-over frequency specifications. Using
the formulae/plots on slide 6/7 of Part IV, the required damping and natural frequency are
ψ ≈ 0.44 and ωn ≈ 3/0.82 = 3.66 rad/sec, respectively. The real part of the conjugate poles
that have these characteristics is −ψωn = −1.61. As such, all of the performance specifications
collectively require the characteristic polynomial to be

γ(s + 10 × 1.61)(s2 + (2 × 1.61)s + 3.662 ) = s2 (τp s + 1) + K(τz s + 1).

Equating the co-efficients of the different powers of s yields the following four equations in the
four unknowns γ, K, τp , τz :

τp = γ; 1 = γ(16.1 + 2 × 1.61); Kτz = γ(16.1 × 2 × 1.61 + 3.662 ); K = γ × 16.1 × 3.662 .

Problem 2
The following number are approximate, as read from the plots by eye.

1. F must be less than 10−24/20 ; see Figure 5. For F ≈ 10−24/20 the open loop-gain is greater than
1 up to around 100 rad/sec and the phase margin is approximately 90◦ .

2. F must be less than 10−12/20 ; see Figure 6. For F ≈ 10−12/20 the open loop-gain is greater than
1 up to around 25 rad/sec and the phase margin is approximately 90◦ .

3. F must be less than 1; see Figure 7. For F ≈ 1, the open loop-gain is greater than 1 up to
around 620 rad/sec and the phase margin is approximately 90◦ .

5
 Bode Diagram
45
40 F = 0.01
35
30
25 F = AN S
20
Magnitude (dB)

15
√
10 F = 0.1
5
0 F =1
−5
−10
−15
−20
−25
−30
0
−10
−20
−30
−40
−50
−60
Phase (deg)

−70
−80
−90
−100
−110
−120
−130
−140
−150
−160
−170
−180 0 1 2 3 4
10 10 10 10 10
Frequency (rad/s)

Figure 5: Bode plots Problem 2 Part 1

6
 Bode Diagram
45
40
35
30
25
20
Magnitude (dB)

15
10 F = AN S
5
0 F =1
−5
−10
−15
−20
−25
−30
−35
0
−10
−20
−30
−40
−50
−60
Phase (deg)

−70
−80
−90
−100
−110
−120
−130
−140
−150
−160
−170
−180 0 1 2 3 4
10 10 10 10 10
Frequency (rad/s)

Figure 6: Bode plots Problem 2 Part 2

7
 Bode Diagram
45
40
35
30
25
20
Magnitude (dB)

15
10
5
0 F = AN S
−5
−10
−15
−20
−25
−30
0
−10
−20
−30
−40
−50
−60
Phase (deg)

−70
−80
−90
−100
−110
−120
−130
−140
−150
−160
−170
−180 0 1 2 3 4
10 10 10 10 10
Frequency (rad/s)

Figure 7: Bode plots Problem 2 Part 3