Homework #6

Prof. M’Closkey

Due Date
The homework is due by 11pm, May 11, 2021. Upload a PDF to the link provided on CCLE.

Problem 1
Analysis of a system zero. Consider the system from Homework #5,

M = Kvin
i Ω1 disks rub
with friction
Ω2
K
+ M
vin
− J J
b

The flywheel inertia J, and the friction coefficient, b, are the values determined in Homework #4,
and K = 1. Use these values in all calculations. The flywheel angular velocities, denoted Ω1 and
Ω2 , are the dependent variables –the flywheel angles are not of interest.
1. What is the value of the zero in the transfer function from vin to Ω1 ?
2. Determine vin for t ≥ 0, and initial flywheel angular velocities Ω1 (0) and Ω2 (0), such that
Ω1 (t) = 0 for t ≥ 0.
3. Show that the net moment on Flywheel #1 is zero at any time with the input and initial
conditions determined in Part 2, thus, if the flywheel is initially at rest, it remains at rest.

Problem 2
Fourier Series analysis of a familiar convolution. Consider the graphical convolution problem
from Problem 1, Homework #3. This problem can be solved using a Fourier series approach since
the input is a periodic signal. Answer the following:
1. Compute the Fourier series coefficients of the periodic signal f .
2. Compute the following partial sum of Fourier series for f on the time grid t=[-3:0.001:3],
200
1 X
f (t) ≈ ck ejkω0 t
T
k=−200

where the ck are the Fourier series coefficients found in Part 1. Note “Gibbs phenomenon”
at points of discontinuity of f (use Matlab to compute and graph the partial sum).

1
 3. The convolution y = g ∗ f is periodic and its Fourier series is given by
∞
1 X
y(t) = ĝ(kω0 )ck ejkω0 t .
T
k=−∞

where ĝ is the Fourier transform of g. Perform a partial sum k = −200, −199, . . . , 199, 200,
on the time grid t=[-3:0.001:3]. Why is does Gibbs phenomenon not occur in the Fourier
series for y (use Matlab to compute and graph the partial sum)? Also compare the Fourier
series analysis result with the graphical convolution you completed in Homework #3.

Problem 3
Fourier transform analysis of a multi-input system. Consider the system shown below
tachometer
⌦
M
J v va
vb
v R1
vb
b 1 1 +
2R R C1 u
R y

There are two inputs to this system: M is a torque applied to the flywheel (in Newton-meters,
denoted N-m), and u is a voltage applied to the low-pass filter (in volts). The output/dependent
variable of interest is the voltage y. The tachometer converts the lywheel angular velocity into the
voltage va such that va = Ω. Here are the parameter values:

J = 0.25 kg-m2 , b = 1 N-m/rad/s, R1 = 1000 ohms, C1 = 0.001 F

2
Answer the following:

1. Complete the block diagram shown below with appropriate frequency responses/transfer func-
tions or gains:

tachometer

⌦ va
M 1 y

vb vb
u 1

isolation amplifier

There are two transfer functions relating to the dependent variable y: one for each input.
The transfer functions are often denoted y/M and y/u to indicate the dependent variable
(y in this case) and the input (M and u). Using the fact that transfer functions/frequency
responses multiply when connected in series, use the block diagram to determine y/M and
y/u.

2. Suppose u = 0 but that M is given by

M = e−t µ(t).

The systems are initially at rest at t → −∞ so there is no zero-input response to consider.

The zero-state response of y is given by

y = hM ∗ M + hu ∗ u,

where hM represents the y/M impulse response and hu represents the y/u impulse response.
The Fourier transform of M does exist and so do the Fourier transforms of the circuit and
flywheel impulse responses. This means Fourier transforms can be used to analyze all of the
signals,
ŷ = ĥM M̂ + ĥu û.
Compute the Fourier transform of M and use it to determine ŷ when u = 0. A partial fraction
expansion of ŷ can be used to separate the expression into readily recognizable functions. Do
this calculation to find y (a function of time).

3. Now suppose you want to “cancel” the effect of M on y using u 6= 0. The objective is to
determine an analytical expression for û so that ŷ = 0 since ŷ = 0 ⇐⇒ y(t) = 0. In this case
0
ŷ



 = ĥM M̂ + ĥu û.

3
 so an expression for û can be determined. Invert the result for û to determine u as a function
of time. Why is this result not surprising? (Observe that M = e−t µ(t) is actually the impulse
response of the circuit, therefore, if u “looks like” the impulse response of the flywheel, then
the effects of M and u on y can cancel each other...)
4. Now find u so that y(t) = e−t µ(t) (still assuming M = e−t µ(t)). First compute û and then
determine u as a function of time.

Problem 4
IVP solution for a more complicated system. Consider the following three flywheel system:

i M = Kvin Ω1 disks rub

with friction
Ω2 disks rub
with friction
Ω3
+ K M
vin
−
J1 J2 J3
b4 b2 b5 b3
b1
attached to J1 attached to J2 attached toJ3

The three wheels with inertias J1 , J2 and J3 spin independently but are coupled via friction from
rubbing disks. The torque created by the disk friction is equal to the frictional constant (either b4
or b5 ) times the difference in angular velocities of the disks (determined by which wheels they are
attached to). The wheels also experience bearing friction modeled by the constants b1 , b2 and b3 .
The angular velocities of the wheels are denoted Ω1 , Ω2 , and Ω3 . The input is the voltage denoted
vin which drives a motor/amplifier combination that provides the torque M to the shaft attached
to the left-most wheel. Answer the following:
1. Derive the equations of motion for this system and write them in the following form:
J1 Ω̇1 = ? Ω1 + ? Ω2 + ? Ω3 + ? vin
J2 Ω̇2 = ? Ω1 + ? Ω2 + ? Ω3 + ? vin
J3 Ω̇3 = ? Ω1 + ? Ω2 + ? Ω3 + ? vin ,
where you need to determine the “?” coefficients (some may be 0).
2. For the remainder of the questions, set all parameters to “1”, i.e. J1 = J2 = J3 = 1, K = 1,
b1 = b2 = b3 = b4 = b5 = 1.

Determine the transfer functions associated with the dependent variables by setting vin = est
and assuming Ω1 = H1 est , Ω2 = H2 est and Ω3 = H3 est , where H1 , H2 and H3 represent the
transfer functions. Rearrange the relations into the following form,
    
? ? ? H1 ?
? ? ? H2  = ?
? ? ? H3 ?
| {z }
3×3 matrix

4
 where the “?” are to be determined. It’s not necessary to invert the 3 × 3 matrix since
substitution can be used to find H2 first.

3. Based on the transfer functions, determine the impulse response associated with Ω2 .

4. Let vin = e−t , t ≥ 0. Find the zero-state response associated with Ω2 . This can be done
by time-domain convolution or, alternatively, by using the Fourier transform since all time
signals are transformable.

5. Since the system is governed by three, coupled, first-order ODEs, each dependent variable is
ultimately governed by a third-order ODE. The determinant of the 3×3 matrix from Part 2 is
the characteristic polynomial of the system. Compute this determinant and the characteristic
roots. Denote the characteristic roots {λ1 , λ2 , λ3 }. Show that the system is asymptotically
stable because all characteristic roots have real part strictly less than 0.

6. Consider the initial conditions for the zero-input response problem,

Ω1 (0) = 1
Ω2 (0) = −1 (1)
Ω3 (0) = 1

Compute the zero-input response associated with Ω2 . Note that zero-input response for any
of the dependent variables is given by

Aeλ1 t + Beλ2 t + Ceλ3 t , t≥0

where the constants {A, B, C} are determined by enforcing the appropriate initial conditions.
The initial conditions for an Ω2 IVP are {Ω2 (0), Ω̇2 (0), Ω̈2 (0)}, where Ω̇2 (0) and Ω̈2 (0) are
determined from the original first-order ODEs and the natural initial conditions (1).