ELEN90055 Control Systems

The University of Melbourne

Problem Set 3 with Reference Solutions
Prepared by M. Cantoni, April 2012

Problem 1
Consider the DC motor servo (position control) system shown in Figure 1. This involves the feedback
.
of a linear combination of the angular velocity ω = θ̇ (measured by a tachometer) and the angular
. .
position θ. Note that Θ(s) = L[θ](s) and Ω(s) = L[ω](s) = sΘ(s) (assuming zero initial conditions),
where L denotes the Laplace transform. The controller gains K and α are real non-negative constants
and Θref (s) denotes the Laplace transform of the reference signal corresponding to the desired angular
position trajectory across time.

Θref (s) + 1 Ω(s) 1 Θ(s)

K
- 0.5s + 1 s

+
+

Figure 1: DC motor servo system (with velocity feedback).

Θ(s)
1. Derive an expression for the transfer function T (s) = .
Θref (s)
2. With α = 0, generate the locus of the poles of T (s) as K varies from 0 to ∞.
3. With α = 1, generate the locus the poles of T (s) as K varies for 0 to ∞.
4. With α = 0.2, generate the locus of the poles of T (s) as K varies from 0 to ∞.
5. With K = 1, generate the locus of the Poles of T (s) as α varies from 0 to ∞.
6. Using the four plots generated in parts 2–5, discuss the benefits of velocity feedback in terms
of achievable damping and speed of the step reference response.

Problem 2
For many aircraft, such as the one illustrated in Figure 2, the pitch angle θ is controlled via the
elevator deflection δ. In the vicinity of a reference pitch angle θ0 > 0 the pitch rate ω = θ̇ is related
(in the frequency domain) to elevator deflection by the transfer function
s+1
G2 (s) = .
(s + 3)(s − 2)

1
The elevator is driven by a hydraulic servo actuator with transfer function
1
G1 (s) =
0.1s + 1
from the elevator command δc to elevator deflection.

Figure 2: Aircraft with tail deflectors for pitch control.

1. What is the response of the pitch rate to a (small) step change of the elevator command?
2. Figure 3 shows the block diagram of a pitch rate control system, based on gyroscopic measure-
ment of the pitch rate ω = θ̇, where K is a constant gain, Ωref (s) is the Laplace transform
. .
of the pitch-rate reference signal generated by the pilot, ∆c (s) = L[δc ](s), ∆(s) = L[δ](s) and
.
Ω(s) = L[ω](s).

Ωref (s) + ∆c (s) ∆(s) Ω(s)

K G1 (s) G2 (s)
-

Figure 3: Pitch rate control system.

(a) For what values of the controller gain K is the closed-loop system stable?
(b) Use a root-locus plot to determine if it is possible to choose K so that all closed-loop poles
are real valued.

Problem 3
1. The left side of Figure 4 shows the Nyquist plot of a plant transfer function. Which of the plots
shown in Figure 5 is the locus of the roots of the closed-loop characteristic polynomial for a
constant gain controller?
2. The right side of Figure 4 shows the Nyquist plot of a plant transfer function. Which of the
plots shown in Figure 6 is the locus of the roots of the closed-loop characteristic polynomial for
a constant gain controller?

2
 Nyquist Diagram

Nyquist Diagram

1.5

0.4

0.5 0.2
Imaginary Axis

Imaginary Axis
0
0

−0.5

−0.2

−1

−0.4
−1.5

−2 −0.6
−1 −0.5 0 0.5 1 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
Real Axis Real Axis

Plot for Problem 3.1 Plot for Problem 3.2

Figure 4: Nyquist Plots for Problem 3.

Root Locus Root Locus Root Locus

1 1 1

0.8 0.8 0.8

0.6 0.6 0.6

Imaginary Axis (seconds−1)

Imaginary Axis (seconds )

Imaginary Axis (seconds )

−1

−1

0.4 0.4 0.4

0.2 0.2 0.2

0 0 0

−0.2 −0.2 −0.2

−0.4 −0.4 −0.4

−0.6 −0.6 −0.6

−0.8 −0.8 −0.8

−1 −1 −1
−1.2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2
Real Axis (seconds−1) Real Axis (seconds−1) Real Axis (seconds−1)

RLOCUS A RLOCUS B RLOCUS C

Figure 5: Root-locus Plots for Problem 3.1.

Root Locus Root Locus Root Locus

0.25 0.25 1

0.2 0.2 0.8

0.15 0.15 0.6

Imaginary Axis (seconds )

Imaginary Axis (seconds−1)

−1

Imaginary Axis (seconds )

−1

0.1 0.1 0.4

0.05 0.05 0.2

0 0 0

−0.05 −0.05 −0.2

−0.1 −0.1 −0.4

−0.15 −0.15 −0.6

−0.2 −0.2 −0.8

−0.25 −0.25 −1
−0.1 −0.08 −0.06 −0.04 −0.02 0 0.02 −0.1 −0.05 0 0.05 0.1 −1 −0.8 −0.6 −0.4 −0.2 0 0.2
Real Axis (seconds−1) Real Axis (seconds−1) Real Axis (seconds−1)

RLOCUS A RLOCUS B RLOCUS C

Figure 6: Root-locus Plots for Problem 3.2.

3
Reference Solutions:

Problem 1
1. From the block diagram in Figure 1,
1 1 K 1 K
Θ(s) = Ω(s) = (Θref (s) − αΩ(s) − Θ(s)) = (Θref (s) − (αs + 1)Θ(s))
s s (0.5s + 1) s (0.5s + 1)

Θ(s) K
whereby = . The poles of T (s) coincide with the roots of the
Θref (s) s(0.5s + 1) + K(αs + 1)
characteristic polynomial
φ(s) = s(0.5s + 1) + K(αs + 1).
Note φ(s) = 0 is equivalent to
αs + 1
1+K =0
s(0.5s + 1)
provided s 6= 0 and s 6= −2 (observe that s = 0 and s = −2 are the roots of φ(s) when K = 0).

Root Locus Root Locus

4 4

3 3
Imaginary Axis (seconds )

Imaginary Axis (seconds )

−1

−1

2 2

1 1

0 0

−1 −1

−2 −2

−3 −3

−4 −4
−10 −8 −6 −4 −2 0 2 −10 −8 −6 −4 −2 0 2
Real Axis (seconds−1) Real Axis (seconds−1)

Root-locus α = 0 Root-locus α = 1

Root Locus Root Locus

1
3
Imaginary Axis (seconds )

Imaginary Axis (seconds−1)

−1

2
0.5
1

0 0

−1
−0.5
−2

−3
−1

−4
−10 −8 −6 −4 −2 0 2 −3 −2.5 −2 −1.5 −1 −0.5 0 0.5
−1 −1
Real Axis (seconds ) Real Axis (seconds )

Root-locus α = 0.2 Root-locus K = 1

Figure 7: Root-locus Plots for Problem 1

4
2. As shown in the north-west corner of Figure 7, the root-locus of φ(s) for α = 0:
1
• starts at the poles of s(0.5s+1) – i.e. at s = 0 and s = −2;
1
• ends at the zeros of s(0.5s+1) – there are no finite zeros;
• lies on the real axis lies to the left of an odd number of real poles and zeros – i.e. between
s = 0 + j0 and s = −2 + j0; and
1
• since s(0.5s+1) has relative degree n − m = 2 − 0, as K → ∞ the locus of 2 roots approaches
a value with infinite magnitude, Palong asymptotes
P at angles η = π/2, 3π/2 radians, which
intersect the real axis at σ = ( poles − zeros)/(n − m) = (−2 + 0)/2 = −1.
√
For α = 0, the characteristic polynomial is 0.5s2 + s + K. This has roots at s = −1 ± 1 − 2K,
which move in the complex plane as shown in the north-west plot in Figure 7 when K is varied
from 0 to ∞. For K = 0.5 the roots meet at s = −1. For K > 0.5 the roots have non-zero
complex part that increases as K does and the real part remains at −1.

3. As shown in the north-east corner of Figure 7, the root-locus of φ(s) for α = 1:

(s+1)
• starts at the poles of s(0.5s+1) – i.e. at s = 0 and s = −2;
(s+1)
• ends at the zeros of s(0.5s+1) – i.e. at s = −1;
• lies on the real axis lies to the left of an odd number of real poles and zeros – i.e. between
s = 0 + j0 and s = −1 + j0 and then between s = −2 + j0 and s = limσ→−∞ σ + j0; and
(s+1)
• since s(0.5s+1) has relative degree n − m = 2 − 1 = 1, as K → ∞ the locus of 1 root
approaches a value with infinite magnitude, along an asymptote at angle η = π/1 radians.

4. As shown in the south-west corner of Figure 7, the root-locus of φ(s) for α = 0.2:
(0.2s+1)
• starts at the poles of s(0.5s+1) – i.e. at s = 0 and s = −2;
(0.2s+1)
• ends at the zeros of s(0.5s+1) – i.e. at s = −5;
• lies on the real axis lies to the left of an odd number of real poles and zeros – i.e. between
s = 0 + j0 and s = −2 + j0 and then between s = −5 + j0 and s = limσ→−∞ σ + j0;
(0.2s+1)
• since s(0.5s+1) has relative degree n − m = 2 − 1 = 1, as K → ∞ the locus of 1 root
approaches a value with infinite magnitude, along an asymptote at angle η = π/1 radians.

To find where the roots meet on the real axis, consider the problem of finding real values K > 0,
k and r such that

φ(s) = s(0.5s + 1) + K(0.2s + 1) = 0.5s2 + (1 + 0.2K)s + K

= k(s + r)2 = ks2 + 2krs + kr2 .

Matching coefficients of sn (n = 0, 1, 2) yields the following three equations:

k = 0.5; (1 + 0.2K) = 2kr; kr2 = K. (1)

As such, K = 0.5r2 and K = 5r − 5. Substituting

√ the latter into the former yields
√ the quadratic
2
equation 0.5r − 5r + 5 = 0; that is, r = (5 ± 25 − 4 · 0.5 · 5)/(2 · 0.5) = 5 ± 15.

5. First note that for K = 1 the characteristic polynomial becomes φ(s) = s(0.5s + 1) + (αs + 1).
In this case φ(s) = 0 is equivalent to
s
1+α =0
0.5s2 +s+1

5
 provided s 6= −1 ± j (note that s = −1 ± j are the characteristic roots for α = 0).
As shown in the south-east corner of Figure 7, the root-locus of φ(s) for K = 1:

• starts at the poles of s

0.5s2 +s+1
– i.e. at s = −1 ± j;
• ends at the zeros of s
0.5s2 +s+1
– i.e. at s = 0;
• lies on the real axis lies to the left of an odd number of real poles and zeros – i.e. between
s = 0 + j0 and s = limσ→−∞ σ + j0; and
• since 0.5s2s+s+1 has relative degree n − m = 2 − 1 = 1, as K → ∞ the locus of 1 root
approaches a value with infinite magnitude, along an asymptote at angle η = π/1 radians.

6. Without velocity feedback (i.e. α = 0), increasing the controller gain K from zero initially
tends to make the closed-loop response have a faster settling time, by moving the slowest pole
further away from the origin s = 0. See the north-west corner of Figure 7. Beyond a certain
point, however, the real part of the closed-loop poles remains fixed at s = −1. So increasing
the gain further yields no reduction in the settling time. Rather, increasing K now reduces the
damping as the poles become complex, with increasing imaginary part; this makes the angle
to the negative real axis increase with K. Reducing the damping leads to more oscillatory
transients (i.e. the decaying natural part) in the step response.
With a small amount of velocity feedback (α = 0.2), increasing K from 0 has the same effect as
when there is no velocity feedback initially. See the south-west corner of Figure 7. Continuing
to increase K, however, now results in the closed-loop poles moving even further away from
the imaginary axis, initially with
√ damping reduction and then finally damping increase. The
roots meet again at s = −5 − 15, with a corresponding fast settling time and no oscillations
in the transient response. With further increase in K, one closed-loop pole moves back towards
s = −5 + j0, leading to a reduction in settling time relative to the fastest possible.
Increasing the velocity feedback gain α too much prevents the potential speed of response
benefits just described. From the north-east plot in Figure 7, it can be seen that for α = 1, the
location of the zero at s = −1 prevents the poles from meeting. In particular, the dominant
pole cannot move to the left of s = −1. The presence of the zero also forces the faster pole to
become even faster (and hence less relevant) while remaining real valued as K increases. For
α = 1, increasing the controller gain K never leads to oscillations in the step response as the
poles are always real valued; but the speed of the response (i.e. 1/settling-time) is limited.
From the south-east corner of Figure 7, it can be seen that increasing α from 0 (with K
fixed) leads to increased damping (as the angle of the complex poles to the negative real axis
decreases). This is consistent with the observations made above. However, beyond a certain
point, increasing α draws one closed-loop pole ever closer to the origin s = 0, thereby reducing
the speed of response (i.e the time it take for the natural components to decay).

BEWARE: The preceding analysis is not the whole story, because it only considers the complemen-
tary sensitivity function. What about the other sensitivities?!?

Problem 2
1. The pitch rate response to a step of size δ̂ in the elevator deflection is given by

δ̂ (s + 1) 1 δ̂
Ω(s) = G2 (s) · G1 (s) · = · ·
s (s + 3)(s − 2) 0.1s + 1 s
in the frequency domain. Note in particular that the largest right-half plane over which this
function is bounded is ℜe(s) > 2. This lies entirely to the right of the imaginary axis (i.e. {s :

6
 Root Locus

10

8 Asymptote
η = π/2
6

Imaginary Axis (seconds−1)

4

−2

−4

−6 Asymptote
η = 3π/2
−8

−10
−10 −8 −6 −4 −2 0 2
−1
Real Axis (seconds )

Figure 8: Root-locus Plots for Problem 2.2(b).

ℜe(s) = 0}). As such, the pitch-rate response signal increases (i.e. grows) exponentially as a
function of time. Indeed, the transfer function G2 (s) is unstable, with a pole at s = +2. This
makes it difficult for a pilot to control directly.

2. Let G(s) = G1 (s)G2 (s) = B(s)/A(s) and C(s) = K = P (s)/L(s), where K > 0 is a constant
gain. The closed-loop characteristic polynomial is given by

φ(s) = A(s)L(s) + B(s)P (s) = (s + 3)(s − 2)(0.1s + 1) + K(s + 1)

= 0.1s3 + 1.1s2 + (K + 0.4)s + (K − 6).

(a) It follows that a necessary condition for the closed-loop poles to be stable is {K − 6 > 0
AND K + 0.4 > 0}. This is equivalent to K > 6, which makes the constraint K > −0.4
redundant. Since the characteristic is a polynomial of degree 3, this condition may not be
sufficient. To this end, consider the corresponding Routh Array:

s3 0.1 K + 0.4
s2 1.1 K −6
1
s (−K − 1.04)/(−1.1) 0
s0 K −6

Using this, it follows that for closed-loop stability it is necessary and sufficient for {K −6 >
0 AND K + 1.04 > 0}, which is equivalent to K > 6 (which just happens to be the same
as the necessary condition previously identified).
(b) As shown in Figure 8, the root-locus of φ(s) with respect to K:
• starts at the poles of (s+3)(s−2)(0.1s+1) – i.e. at s = −10, −3, +2;
s+1

• ends at the zeros of (s+3)(s−2)(0.1s+1) – i.e. at s = −1;

s+1

• lies on the real axis lies to the left of an odd number of real poles and zeros – i.e. between
s = 2 + j0 and s = −1 and between s = −3 + j0 and s = −10 + j0; and
• since (s+3)(s−2)(0.1s+1)
s+1
has relative degree n − m = 3 − 1 = 2, as K → ∞ the locus
of 2 roots approaches a value with infinite magnitude,Palong an P asymptote at angle
η = π/2, 3π/2 radians, which intersect the real axis at ( poles − zeros)/(n − m) =
(−10 − 3 + 2 − (−1))/2 = −5.

7
 It seems possible, given sufficiently small values of K, for the closed-loop poles (roots of
the characteristic polynomial) to be all real; however, these value of K may not be large
enough to achieve closed-loop stability (the locus of one root starts at s = 2 + j0 in the
right-half plane). Recall from above that for closed-loop stability we need K > 6. Another
way to see this is to determine the value of K for which the root-locus coincides with the
imaginary axis. This happens for values of ω such that φ(jω) = 0; that is,

−j0.1ω 3 − 1.1ω 2 + j(K + 0.4)ω + (K − 6) = 0,

which (by matching real and imaginary parts) gives rise to the pair of equations −11ω 2 +
K − 6 = 0 and −ω(0.1ω 2 − (K + 0.4)) = 0, which have simultaneous solution at (ω =
0, K = 6). Indeed, when K = 6 the characteristic √ polynomial is s(0.1s2 − 1.1s + 6.4)
with a pole at s = 0 and poles at s = −5.5 ± 5 1.21 − 0.4 · 6.4 (which are complex since
0.4 · 6.4 > 1.21). To summarise, it is not possible to choose a gain that is both stabilizing
and such that the closed-loop poles are all real valued. Therefore, there will be some
oscillation in the transient response to step changes in the reference pitch rate. Increasing
K leads to decreased damping of the transient as the complex poles move towards infinity
along the asymptotes.

Problem 3
1. The Nyquist plot for Problem 3.1 in Figure 4 matches the root-locus plot RLOCUS B in Figure 5.
All three root-locus plots correspond to plant transfer functions that have relative degree equal
to 2. This is consistent with the Nyquist plot moving to the origin as ω → ∞. It is also
consistent with the origin being approached at an angle of π radians (zoom in).
It can’t be RLOCUS A for the following reasons:

• From the uncompensated Nyquist plot it is apparent that increasing the controller gain K
will eventually lead to 2 anti-clockwise encirclements of the critical point s = −1. This is
not possible for an open-loop transfer function with only one unstable pole; it would imply
a negative number of unstable closed-loop poles which doesn’t make sense.
• For a transfer function with RLOCUS A (i.e. 1 fast stable pole, 1 slow unstable pole, no
zeros), for positive frequencies, the phase varies between −π and −π/2 (if the DC gain is
negative) and 0 and π/2 (if the DC gain is positive). This is not the case for the Nyquist
plot shown, where for positive frequencies the phase varies between 0 and 3π/2.

Similarly, it cannot be RLOCUS C. As noted above, for sufficiently large controller gain K, the
compensated Nyquist plot makes 2 anti-clockwise encirclements of the critical point s = −1.
According to RLOCUS C the open-loop plant has no unstable poles. Therefore, 2 anti-clockwise
encirclements would imply a negative number of closed-loop poles, which doesn’t make sense.
This leaves RLOCUS B. The number of encirclements made by the compensated Nyquist plot is
consistent with the number of unstable open-loop poles and closed-loop stability for sufficiently
high gains.

2. The Nyquist plot for Problem 3.2 in Figure 4 matches the root-locus plot RLOCUS A in
Figure 6.
From the Nyquist plot it is apparent that as the controller gain is increased from K = 0, the
number of clockwise encirclements changes from 0 to 2 and then back to 0. For an open-loop
transfer function with all stable poles (which is the case for all root-locus plots given), this
would be consistent with the closed-loop being stable for small gains, unstable with two poles

8
in the right-half plane for an intermediate range of gains, and stable for large gains. The only
root-locus plot that exhibits such behaviour is RLOCUS A.
It can’t be RLOCUS B because for high-gain control the closed-loop is unstable, yet there are
no clockwise encirclements of the critical point made by the Nyquist plot for very large K.
It can’t be RLOCUS C for a similar reason. Another reason is that the relative degree is 2
in this case, which is inconsistent with the phase with which the Nyquist plot approaches the
origin for large frequencies; this is consistent with a transfer function with relative degree 1.