6.

003 Homework #6 Solutions

Problems
1. Maximum gain
For each of the following systems, find the frequency ωm for which the magnitude of the
gain is greatest.
1
a.
1 + s + s2

q
ωm = 1
2

√
3
This system has poles at s = − 12 ± j 2 .

√
3
2
ω
− 12

√
− 23

We must minimize the product of the lengths of the two vectors. The product of the
squared lengths is
 √ 2  2 !  √ 2  2 !
3 1 3 1
ω− + ω+ + = 1 − ω2 + ω4
2 2 2 2

Minimize by taking derivative with respect to ω and setting to zero:

d
= −2ω + 4ω 3 = 2ω(2ω 2 − 1) = 0
dω
The solutions are ω
q= 0 (which corresponds to a local maximum in the product of the
lengths) or ω = ± 12 . Thus the desired solution is
r
1
ω= .
2
6.003 Homework #6 Solutions / Fall 2011 2
s
b.
1 + s + s2

ωm = 1

Now there is an added vector from the zero at s = 0.

√
3
2
ω
− 12

√
− 23

The vector associated with the zero is in the numerator, while those associated with
the poles are in the denominator. The squared quotient of lengths is
ω2
=
1 − ω2 + ω4
Maximize by taking derivative with respect to ω and setting to zero:
d 2ω(ω 4 − 1)
= =0
dω (1 − ω + ω 4 )2
The solutions are ω = 0 (which corresponds to a minimum of the magnitude of the
gain) or ω = ±1 or ω = ±j. The desired solution is real and positive:

ω = 1.
6.003 Homework #6 Solutions / Fall 2011 3
s2
c.
1 + s + s2

√
ωm = 2

Now there are two added vectors from the zeros at s = 0. The squared quotient of
lengths is
ω4
=
1 − ω2 + ω4
Maximize by taking derivative with respect to ω and setting to zero:
d 2ω 3 (2 − ω 2 )
= =0
dω (1 − ω + ω 4 )2
√
The solutions are ω = 0 (which again corresponds to a minimum) or ω = ± 2. The
desired solution is
√
ω = 2.

Compare the ωm for these systems and make sure that you can explain qualitatively any
similarities or differences.
The frequency of maximum gain increased from part a to b to c. The increase is because
of the zero in part b and because of the two zeros in part c. The effect of the zero is
to weight the gain by a factor that grows with frequency. This added weight pushes the
maximum gain to progressively higher frequencies.
6.003 Homework #6 Solutions / Fall 2011 4
2. Phase
For a second-order system with poles at −1 and −4 (and no zeros), find the frequency at
which the phase is −90◦ , using any method except for the vector method. Then illustrate
and confirm that result using the vector method.

ω= 2

The system function has the form

C C
H(s) = = 2 .
(s + 1)(s + 4) s + 5s + 4
Thus, the frequency response has the form
C C
H(jω) = = 2
.
(jω + 1)(jω + 4) 4 − ω + j5ω
For very small values of ω, the denominator of H(jω) is a small real number. As ω
increases, the real part of the denominator decreases and the imaginary part increases.
Assuming that C is positive and real, H(jω) can only have an angle of −90◦ if the denom-
inator has an angle of 90◦ — which means that the real part of the denominator must be
zero:
ω2 = 4 .

It follows that ω = 2.
Next we illustrate this result using the vector method. The proposed −90◦ frequency is
ω = 2, so put in s = jω0 = 2j. There, the phase of (s + 1)(s + 4) is the sum of the two
angles θ1 and θ2 in this figure:
Im

ω=2

θ2
θ1
Re
−4 −1
The large, enclosing triangle (thick line) is a right triangle with side ratio 4 : 2. The
smallest triangle is a right triangle with side ratio 1 : 2, which is the reverse of the side
ratio for the large triangle. So the small triangle has the same shape as the large triangle
but is rotated by 90 degrees (and is shrunk by a factor of 2). Thus θ1 and θ2 are the two
non-right-angle angles of either triangle, and θ1 + θ2 = 90◦ . So the denominator has a 90◦
phase at ω = 2, meaning that the overall system has a −90◦ phase at ω = ω0 , which is
ω = 2 as we wanted to show.
6.003 Homework #6 Solutions / Fall 2011 5
3. CT stability
Consider the following feedback system in which the box represents a causal LTI CT
system that is represented by its system function.

K
X + Y
− s2 +s−2

The closed-loop system response is

K
s2 +s−2 K
H(s) = K
= .
1 + s2 +s−2 s2 + s + (K − 2)

The closed-loop poles are at

r
1 1
s=− ± +2−K
2 4
as shown in the following figure, for K > 0.

s-plane

−2 1

a. Determine the range of K for which this feedback system is stable.

range of K: K > 2

This system is stable when the closed-loop poles are in the left half-plane, which is
true for all K > 2.

b. Determine the range of K for which this feedback system has real-valued poles.

9
range of K: −∞ < K < 4

9
This system is has real valued poles if −∞ < K < .
4
6.003 Homework #6 Solutions / Fall 2011 6
4. DT stability
Consider the following feedback system in which the box represents a causal LTI DT
system that is represented by its system function.

K
X + Y
− z2 +z−2

The closed-loop system response is

K
z 2 +z−2 K
H(z) = K
= .
1 + z 2 +z−2 z2 + z + (K − 2)

The closed-loop poles are at

r
1 1
z=− ± +2−K
2 4
as shown in the following figure, for K > 0.

z -plane

−2 1

a. Determine the range of K for which this feedback system is stable.

range of K: 2 < K < 3

This system is stable when the closed-loop poles are inside the unit circle. To get the
left pole inside the unit circle, K must be bigger than 2. To keep both complex-valued
poles inside the unit circle, K < 3. In total 2 < K < 3.

b. Determine the range of K for which this feedback system has real-valued poles.

9
range of K: −∞ < K < 4

9
This system is has real valued poles if −∞ < K < .
4
6.003 Homework #6 Solutions / Fall 2011 7
Engineering Design Problems
5. Automotive suspension
Wheels are attached to an automobile through a suspension system that is designed to
minimize the vibrations of the passenger compartment that result when traveling over
bumpy terrain. The suspension system consists of a spring and shock absorber that
are both compressed when the wheel passes over a bump, so that the sudden motion
of the wheel is not directly transmitted to the passenger compartment. The spring
generates a force to hold the passenger compartment at a desired distance above the
surface of the road, and the shock absorber adds frictional damping. In this problem,
you will determine how much damping is desireable by analyzing a simple model of an
automobile’s suspension system shown below.
M y(t)

K B

x(t)

The model consists of a mass M that represents the mass of the car, which is connected
through a spring and dashpot to the wheel. The vertical displacement of the wheel from
it’s equilibrium position is taken as the input x(t). The vertical displacement of the
mass from it’s equilibrium position is taken as the output y(t). The spring is assumed to
obey Hooke’s law, so that the force it generates is a constant K times the amount that
the spring is compressed relative to it’s equilibrium compression. The shock absorber
is assumed to generate a force that is a constant B times the velocity with which the
shock absorber is compressed. Notice that by referring x(t) and y(t) to their equilibrium
positions, the force due to gravity can be ignored. Assume that M = 1 and K = 1.

a. Determine the differential equation that relates the input x(t) and output y(t).
Start with Newton’s law: F = M a where M represents the mass of the car and a
represents the acceleration
of the car. There are two important forces on the car: the


spring force K x(t) − y(t) and the force generated by the dashpot B ẋ(t) − ẏ(t) .
Combining these, we find that



K x(t) − y(t) + B ẋ(t) − ẏ(t) = M ÿ(t)

or equivalently
ÿ(t) + B ẏ(t) + y(t) = x(t) + B ẋ(t)

since M = K = 1.
6.003 Homework #6 Solutions / Fall 2011 8
b. Determine and plot the impulse response of the system when B = 0. Based on this
result, give a physical explanation of the problem that would result if there were no
shock absorber in the system.
If B = 0 then the Laplace transform is:
1 1
1 2j 2j
H(s) = 2
= − .
s +1 s−j s+j
The corresponding impulse response is
1
e jt − e−jt u(t) = sin(t)u(t)


h(t) =
2j
as plotted below.
h(t)
1

Without the shock absorber, there is no energy dissipation, and the car will oscillate
forever after hitting a bump.

c. Determine an expression for the smallest positive damping constant B for which the
poles of the system have real values. Sketch the impulse response of the system for
this value of B. Based on this result, give a physical explanation of how the shock
absorber improves performance of the suspension system.
In general, the Laplace transform is:
Bs + 1
H(s) = ,
s2 + Bs + 1
The poles will have real values when B ≥ 2. When B = 2 there are two poles at
s = −1. Thus
2s + 1 2 1
H(s) = 2 = −
s + 2s + 1 s + 1 (s + 1)2
The corresponding impulse response is
h(t) = (2 − t)e−t u(t).
h(t)
2

The shock absorber improves the system by damping out the oscillations that would
otherwise result.
6.003 Homework #6 Solutions / Fall 2011 9
d. Consider what would happen if B were very large. Sketch the impulse response for
the system if B = 100. Describe how this response might be less desireable than that
in part c. Provide a physical explanation for how a stiff shock absorber can degrade
system performance.
If B = 100 then the poles are approximately s = −100 and s = −0.01,
100s + 1 100s + 1 100
H(s) = 2 ≈ ≈ .
s + 100s + 1 (s + 100)(s + 0.01) s + 100
The corresponding impulse response is
h(t) = 100e−100t u(t) .
h(t)
100

t
The time constant for this response is 10 milliseconds, which is to short to be visible
in the plot above. From the passenger’s point of view, this response is very fast. Since
the integral of h(t) is
Z ∞ Z ∞
h(t)dt = 100e−100t u(t)dt = 1
−∞ −∞

It follows that h(t) is a good approximation to an impulse δ(t). Thus the output y(t)
is nearly equal to the input x(t). The damping is so great that all vibrations of the
wheels are transmitted to the car. The suspension system might just as well not be
there!

6. Dial tones
Pressing the buttons on a touch-tone phone generates tones that are used for dialing.
Each button produces a pair of tones of the form
x(t) = cos(2πf1 t) + cos(2πf2 t)
where f1 and f2 code the row and column of the button as shown in the following table.
f2 [Hz]
f1 [Hz] 1209 1336 1477
697 1 2 3
770 4 5 6
852 7 8 9
941 ∗ 0 #

This problem concerns the design of a system to detect the row and column numbers
that were pressed by analyzing the signal x(t). The following block diagram illustrates
the basic structure of such a system.
y1 [n]
h1 [n] y1 [n] > Γ

y2 [n]
x[n] = x(nT ) h2 [n] y2 [n] > Γ

y3 [n]
h3 [n] y3 [n] > Γ
6.003 Homework #6 Solutions / Fall 2011 10
The input x(t) is first sampled with T = 10−4 seconds. The samples are then passed
through LTI systems that generate intermediate signals so that y1 [n] is large when a
button in column 1 is pressed, y2 [n] is large when a button in column 2 is pressed, and
y3 [n] is large when a button in column 3 is pressed. These intermediate signals are then
passed through detectors that determine when the signals are bigger than a threshold
value Γ. Your task is to design the LTI systems. Each should consist of a system with 2
poles of the form shown in the following pole-zero diagram.
z -plane

r
Ω0

Such systems can be simulated by finding the difference equation that corresponds to the
system and then iteratively solving that difference equation.

a. Determine values of r and Ω0 so that the h1 [n] system generates a large response when
the “1” key is pressed and a small response when the “2” or “’3” keys are pressed.
Your solution should work not only when the input consists of a single key press but
also when it consists of sequences of key presses (as when dialing a phone number).
Submit hardcopies of your code to generate y1 [n] along with a plot of y1 [n].
For this problem consider the following input signal:
exp(jΩ0 n) + exp(−jΩ0 n) 1
x[n] = cos(Ω0 n) = = {exp(jΩ0 )n + exp(−jΩ0 )n } .
2 2
This signal consists of eigenfunctions whose bases are exp(jΩ0 ) and exp(−jΩ0 ). Both
of these are points on the unit circle, with angles given by ±Ω0 .
Consider a system given by two poles at z = r · exp(±jΩ0 ). The frequency response
of this system will have its magnitude peak at frequency ±Ω0 . Hence the signal x[n]
given above will be amplified more when Ω0 ≈ Ω0 . We use this to discriminate the
three possible tones, by setting Ω0 = 2π0.1209.
Given a system with two poles, suppose the two poles are at zA = r exp(jΩ0 ) and
zB = r exp(−jΩ0 ). Then we can write
Y (z) 1 1
= = 2 ,
X(z) (z − zA )(z − zB ) z − (zA + zB )z + zA zB
or in other words, as a difference equation:
z 2 Y (z) − (zA + zB )zY (z) + zA zB Y (z) = X(z),
y[n + 2] − (zA + zB )y[n + 1] + zA zB y[n] = x[n].

r = 0.99;
w_0 = 2*pi*0.1209;
zA = r*exp(j*w_0);
zB = r*exp(-j*w_0);
y = zeros(1,3000);
for n = 1:(length(y)-2)
6.003 Homework #6 Solutions / Fall 2011 11
y(n+2) = (zA+zB)*y(n+1) - (zA*zB)*y(n) + x(n);
end

b. Describe how the choice of Ω0 affects the output signal y1 [n].

The closer Ω0 is to 2π0.1209, the larger the magnitude at the output corresponding
to that tone is. If Ω0 approaches one of the other tones, then the magnitude at the
output corresponding to those respective tones are larger.

c. Describe how the choice of r affects the output signal y1 [n]. In particular, what limits
the maximum acceptable value of r? Also, what limits the minimum acceptable value
of r?
In order to increase the relative amplification of tone 1, we wish to choose r to be close
to 1 but without making the system unstable. Choosing r to be between 0.9 and 0.99
gives good discrimination of the tones without incurring instability. This is shown in
the following figure. When r is chosen to be too close to 1, or greater than 1, the
system becomes unstable.

r = 0.5, w
0
= 2 p 0.1209
2

­2
0 500 1000 1500 2000 2500 3000

r = 0.9, w
0
= 2 p 0.1209
10

­10
0 500 1000 1500 2000 2500 3000

r = 0.99, w
0
= 2 p 0.1209
100

­100
0 500 1000 1500 2000 2500 3000

r = 0.99999, w
0
p 0.1209
= 2

500

­500
0 500 1000 1500 2000 2500 3000