System Dynamics

Problems

Problem 12.1 (TILE)

Use the linear graph below of a thermal system to (a) derive the transfer function \(T_{R_2}(s)/T_s(s)\), where \(T_s\) is the input temperature and \(T_{R_2}\) is the temperature across the thermal resistor \(R_2\). Use impedance methods. And (b) derive the input impedance the input \(T_s\) drives.

A linear graph for problem 12.1.
Figure 12.8: A linear graph for problem 12.1.
  1. The transfer function \(T_{R_2}(s)/T_s(s)\) can be found easily from an across-variable divider rule. The combined impedance of \(R_1\) and \(R_2\) (in parallel, i.e., \(Z_{R_1} \parallel Z_{R_2}\)) retains the temperature difference \(T_{R_2}\), so \[\begin{aligned} \frac{T_{R_2}(s)}{T_s(s)} &= \frac{Z_{R_1} \parallel Z_{R_2}}{(Z_{R_1} \parallel Z_{R_2}) + Z_C } \\ &= \frac{\frac{R_1 R_2} {R_1 + R_2}}{\frac{R_1 R_2} {R_1 + R_2} + \frac{1} {C s}} \\ &= \frac{R_1 R_2 C s} {R_1 R_2 C s + R_1 + R_2}. \end{aligned}\]

  2. The input impedance \(Z_\text{in}\) is the impedance of all the elements combined in series and parallel; that is, \[\begin{aligned} Z_\text{in} &= \left( Z_{R_1} \parallel Z_{R_2} \right) + Z_C \\ &= \frac{R_1 R_2} {R_1 + R_2} + \frac{1} {C s} \\ &= \frac{R_1 R_2 C s + R_1 + R_2} {(R_1 + R_2) C s}. \end{aligned}\]

Problem 12.2 (GRANITE)

Use the linear graph below of a fluid system to (a) derive the transfer function \(P_C(s)/P_S(s)\), where \(P_S\) is the input pressure and \(P_C\) is the pressure across the fluid capacitance \(C\). Use impedance methods and a divider rule is highly recommended. (Simplify the transfer function.) And (b) derive the input impedance the input \(P_S\) drives. (Don’t simplify the expression.)

A linear graph for problem 12.2.
Figure 12.9: A linear graph for problem 12.2.
  1. This is a straightforward across-variable divider: \[\begin{aligned} \frac{P_C(s)} {P_S(s)} &= \frac{Z_C} {Z_I+Z_R+Z_C} \\ &= \frac{1} {I C s^2 + R C s + 1}. \end{aligned}\]

  2. The input impedance for the system consisting of three series elements is \[\begin{aligned} Z_R + Z_I + Z_C = R + I s + \frac{1} {C s}. \end{aligned}\]

Problem 12.3 (GRANTED)

Use the linear graph below of an electronic system to derive the transfer function \(I_{R_1}(s)/V_S(s)\), where \(V_S\) is the input voltage and \(I_{R_1}\) is the current through the resistor \(R_1\). (Simplify the transfer function.) Use an impedance method. Hint: a divider method is recommended; without it, use of a computer is recommended.

A linear graph for problem 12.3.
Figure 12.10: A linear graph for problem 12.3.

We will use an across-variable (voltage) divider. The elements \(C_1\), \(R_2\), and \(C_2\) can all be combined (in series and parallel) to form an equivalent impedance \[\begin{aligned} Z_e &= \frac{1} {\frac{1} {Z_{C_1}} + \frac{1} {Z_{R_2} + Z_{C_2}}} \\ &= \frac{1} {C_1 s + \frac{1} {R_2 + \frac{1} {C_2 s}}} \\ &= \frac{1} {C_1 s + \frac{C_2 s} {R_2 C_2 s + 1}} \\ &= \frac{R_2 C_2 s + 1} {(R_2 C_2 s + 1) C_1 s + C_2 s} \\ &= \frac{R_2 C_2 s + 1} {R_2 C_1 C_2 s^2 + (C_1 + C_2) s}. \end{aligned}\]

The voltage divider rule gives \[\begin{aligned} \frac{V_{R_1}}{V_S} &= \frac{Z_{R_1}}{Z_{R_1} + Z_e} \\ &= \frac{(R_2 C_1 C_2 s^2 + (C_1 + C_2) s) R_1} {(R_2 C_1 C_2 s^2 + (C_1 + C_2) s) R_1 + R_2 C_2 s + 1} \\ &= \frac{R_1 R_2 C_1 C_2 s^2 + R_1 (C_1 + C_2) s} {R_1 R_2 C_1 C_2 s^2 + R_1 (C_1 + C_2) s + R_2 C_2 s + 1} \\ &= \frac{(R_1 R_2 C_1 C_2) s^2 + R_1 (C_1 + C_2) s} {(R_1 R_2 C_1 C_2) s^2 + \left(R_1 (C_1 + C_2) + R_2 C_2\right) s + 1}. \end{aligned}\]

But we want \(I_{R_1}/V_S\). The impedance elemental equation for \(R_1\) lets us trade \(V_{R_1} \mapsto I_{R_1}\): \[\begin{aligned} V_{R_1} &= I_{R_1} Z_{R_1} \\ &= R_1 I_{R_1}. \end{aligned}\] Substituting this into the transfer function above, \[\begin{aligned} \frac{R_1 I_{R_1}}{V_S} &= \frac{(R_1 R_2 C_1 C_2) s^2 + R_1 (C_1 + C_2) s} {(R_1 R_2 C_1 C_2) s^2 + \left(R_1 (C_1 + C_2) + R_2 C_2\right) s + 1} \Rightarrow \\ \frac{I_{R_1}}{V_S} &= \frac{(R_2 C_1 C_2) s^2 + (C_1 + C_2) s} {(R_1 R_2 C_1 C_2) s^2 + \left(R_1 (C_1 + C_2) + R_2 C_2\right) s + 1}. \end{aligned}\]

Problem 12.4 (CONCRETE)

Use the linear graph of a fluid system in to derive the transfer function \(Q_C(s)/P_S(s)\), where \(P_S\) is the input pressure and \(Q_C\) is the flowrate through the fluid capacitance \(C\). Use impedance methods; a divider rule is recommended but not required. Identify all impedances but do not substitute them into the transfer function.

A fluid system linear graph.
Figure 12.11: A fluid system linear graph.

The transfer function can be derived using impedance methods, as required, in two different manners:

  1. with the systematic approach of or

  2. with the divider method of .

We choose the latter because it is a relatively simple system and the across-variable divider looks promising: (a) the source \(P_S\) is across-variable and (b) the linear graph can be reduced to the source \(P_S\) in series with elements that include \(C\), the element associated with the output variable \(Q_C\).

Before we begin with the divider method, let’s list out the impedances associated with each element: \[\begin{gathered} Z_{R_1} = R_1,\ Z_{R_2} = R_2, \\ Z_{I_1} = I_1 s,\ Z_{I_2} = I_2 s,\ Z_C = \frac{1} {C s}. \end{gathered}\]

We must first find an equivalent impedance of the two upper \(R\)-\(I\) “branches.” The fluid resistors combine in series with the inertances as \[\begin{aligned} Z_{R_1} + Z_{I_1} \quad\text{and} Z_{R_2} + Z_{I_2} \end{aligned}\] to form two parallel elements that can be combined to form an element with equivalent impedance \[\begin{aligned} Z_e &= \frac{1} {\frac{1} {Z_{R_1} + Z_{I_1}} + \frac{1} {Z_{R_2} + Z_{I_2}}}. \end{aligned}\] The corresponding simplified graph is shown in .

the simplified linear graph for applying the across-variable divider rule.

We can now apply the across-variable divider rule: \[\begin{aligned} \label{eq:concrete-divider} \frac{P_C(s)} {P_S(s)} &= \frac{Z_C} {Z_C + Z_e}. \end{aligned}\]

This is nearly what we want, \(Q_C/P_S\). For all passive elements, the impedance or “generalized Ohm’s law” elemental equation allows us to easily trade through- and across-variable. The \(P_C\) impedance elemental equation is \[\begin{aligned} P_C = Q_C Z_C. \end{aligned}\] Substituting this into , \[\begin{aligned} \label{eq:concrete-divider2} \frac{Q_C(s) Z_C} {P_S(s)} &= \frac{Z_C} {Z_C + Z_e}. \end{aligned}\] This is simply solved for \[\begin{aligned} \label{eq:concrete-divider3} \frac{Q_C(s)} {P_S(s)} &= \frac{1} {Z_C + Z_e}. \end{aligned}\]

Problem 12.5 (RHINE)

Use the linear graph of a fluid system in figure 12.12 to derive the transfer functions \(Q_C(s)/P_S(s)\) and \(Q_{R_3}(s)/P_S(s)\), where \(P_S\) is the input pressure, \(Q_{R_3}\) is the flowrate through a valve with resistance \(R_3\), and \(Q_C\) is the flowrate through a tank with fluid capacitance \(C\). Use impedance methods; a divider rule is not required. Identify all impedances and substitute them into the transfer functions, but you are not required to simplify these expression.

A fluid system linear graph.
Figure 12.12: A fluid system linear graph.

The transfer function can be derived using impedance methods, as required, in two different manners:

  1. with the systematic approach of or

  2. with the divider method of .

We choose the latter because the across-variable divider looks promising: (a) the source \(P_S\) is across-variable and (b) the linear graph can be reduced to the source \(P_S\) in series with elements that include \(R_3\) and \(C\), the elements associated with the output variables \(Q_{R_3}\) and \(Q_C\).

Before we begin with the divider method, let’s list out the impedances associated with each element: \[\begin{gathered} Z_{R_1} = R_1,\ Z_{R_2} = R_2,\ Z_{R_3} = R_3,\\ Z_{I_1} = I_1 s,\ Z_{I_2} = I_2 s, Z_C = \frac{1} {C s}. \end{gathered}\]

We must first find an equivalent impedance of the two upper \(R\)-\(I\) “branches.” The fluid resistors combine in series with the inertances as \[\begin{aligned} Z_{R_1} + Z_{I_1} \quad\text{and } Z_{R_2} + Z_{I_2} \end{aligned}\] to form two parallel elements that can be combined to form an element with equivalent impedance \[\begin{aligned} Z_e &= \frac{1} {\frac{1} {Z_{R_1} + Z_{I_1}} + \frac{1} {Z_{R_2} + Z_{I_2}}}. \end{aligned}\] The corresponding simplified graph is shown in .

the simplified linear graph for applying the across-variable divider rule.

The across-variable divider also requires we combine the impedances of \(R_3\) and \(C\) in parallel. This is associated with the two output elements, so we call it \(Z_o\), and define it as follows: \[\begin{aligned} Z_o &= \frac{1} {\frac{1} {Z_{R_3}} + \frac{1} {Z_{C}}}. \end{aligned}\]

We can now apply the across-variable divider rule: \[\begin{aligned} \label{eq:rhine-divider} \frac{P_o(s)} {P_S(s)} &= \frac{Z_o} {Z_o + Z_e}. \end{aligned}\]

This is nearly what we want, \(Q_{R_3}/P_S\) and \(Q_C/P_S\). We recognize that the pressure \(P_o = P_{R_3} = P_C\), so we have \(P_{R_3}/P_S\) and \(P_C/P_S\). For all passive elements, the impedance or “generalized Ohm’s law” elemental equation allows us to easily trade through- and across-variable.

The \(P_{R_3}\) impedance elemental equation is \[\begin{aligned} P_{R_3} = Q_{R_3} Z_{R_3}. \end{aligned}\] Substituting this into , \[\begin{aligned} \label{eq:rhine-divider2} \frac{Q_{R_3}(s) Z_{R_3}}{P_S(s)} &= \frac{Z_o} {Z_o + Z_e}. \end{aligned}\] This is simply solved for \[\begin{aligned} \label{eq:rhine-divider3} \frac{Q_{R_3}(s)}{P_S(s)} &= \frac{Z_o/Z_{R_3}}{Z_o + Z_e}. \end{aligned}\]

The \(P_C\) impedance elemental equation is \[\begin{aligned} P_C = Q_C Z_C. \end{aligned}\] Substituting this into , \[\begin{aligned} \label{eq:rhine-divider2} \frac{Q_C(s) Z_C} {P_S(s)} &= \frac{Z_o} {Z_o + Z_e}. \end{aligned}\] This is simply solved for \[\begin{aligned} \label{eq:rhine-divider3} \frac{Q_C(s)} {P_S(s)} &= \frac{Z_o/Z_C} {Z_o + Z_e}. \end{aligned}\]

Problem 12.6 (TABLEAU)

Consider an accelerometer that has transfer function \[\begin{aligned} G(s) \equiv \frac{V_i(s)} {A(s)} = \frac{K_G\,\omega_{n_G}^2}{s^2 + 2 \zeta_G \omega_{n_G}\, s + \omega_{n_G}^2}, \end{aligned}\] where

  • \(A\) is the input acceleration in m/s\(^2\),

  • \(V_i\) is the output voltage in V,

  • \(K_G = 0.1\) V/(m/s\(^2\)) is the gain,

  • \(\omega_{n_G} = 3000\) rad/s is the natural frequency, and

  • \(\zeta_G = 0.2\) is the damping ratio.

Perform a frequency domain analysis as follows.

  1. Generate a Bode plot of \(G(s)\).

  2. At DC (\(\omega = 0\) rad/s), compute the magnitude and phase of the frequency response function of the accelerometer.

Suppose there is a sinusoidal systematic noise signal at the input, with amplitude \(a_\text{noise} = 1\) m/s\(^2\) and frequency \(\omega_\text{noise} = 2900\) rad/s.1

  1. Assuming there is only noise input, at the noise frequency \(\omega_\text{noise}\), compute the amplitude and phase of the voltage \(V_i\) at the output of the accelerometer. Why is the amplitude higher than it would have been at DC (use your Bode plot from to justify your answer).

To mitigate the systematic noise, we add a filter with transfer function \(H(s)\) to the output of the accelerometer, as shown in .

Accelerometer and filter block diagram.

By definition, \[\begin{aligned} H(s) \equiv \frac{V_o(s)} {V_i(s)}. \end{aligned}\] Assume the filter and accelerometer do not dynamically load each other. The filter circuit diagram is shown in .

Filter circuit.
  1. Draw a linear graph model of the filter circuit.

  2. Use impedance methods to derive the transfer function \(H(s)\) in terms of the circuit element parameters \(R\), \(L\), and \(C\).

  3. Find the filter’s natural frequency \(\omega_{n_H}\) and damping ratio \(\zeta_H\).2

  4. Let \(C = 0.001\) F. Design the filter by choosing \(R\) and \(L\) such that \[\begin{aligned} \zeta_H = 1 \quad\text{and}\quad\omega_{n_H} = 1000~\text{rad/s}. \end{aligned}\]

  5. Find the transfer function \[\begin{aligned} \frac{V_o(s)} {A(s)} \end{aligned}\] with all parameters substituted. Simplify.

  6. Generate a Bode plot for \(V_o(s)/A(s)\).

  7. Using the Bode plot of , explain why we should expect the output from the systematic noise at \(\omega_\text{noise}\) to be improved.

  8. From the transfer function \(V_o(s)/A(s)\), at the noise frequency \(\omega_\text{noise}\), compute the amplitude and phase of the output voltage \(V_o\).

  9. Compare the result from to the unfiltered voltage in by finding the ratio of the filtered amplitude over then unfiltered amplitude.

  10. How could you augment the filter design to further reduce the systematic noise?

a: Bode plot

First, define the system with

wn = 3000; % rad/s
z = .2;
K_G = 0.1; % V/(m/s^2)
G = tf([K_G*wn^2],[1,2*z*wn,wn^2])
G =
 
        900000
  -------------------
  s^2 + 1200 s + 9e06
 
Continuous-time transfer function.

Now we can generate the Bode plot, shown in , with

figure;
bode(G);
grid on;

b: DC gain and phase

At DC (\(\omega = 0\) rad/s),

G_DC = freqresp(G,0); % complex FRF at DC
mag_DC = abs(G_DC); % magnitude
pha_DC = angle(G_DC); % phase
fprintf("DC magnitude = %.4g\n",mag_DC)
fprintf("DC phase = %.4g rad",pha_DC)
DC magnitude = 0.1
DC phase = 0 rad

These are the accelerometer DC gain and phase.

c: Noise output

wnoise = 2900; % rad/s ... noise frequency
anoise = 1; % m/s^2 ... noise amplitude

The noise output can be computed as

Gnoise = freqresp(G,wnoise); % FRF at noise freq.
Vinoise_mag = abs(Gnoise)*anoise; % multiplies
Vinoise_pha = 0 + angle(Gnoise);
fprintf("Noise |Vo| = %.4g V\n",Vinoise_mag)
fprintf("Noise ∠Vo = %.4g deg",rad2deg(Vinoise_pha))
Noise |Vo| = 0.255 V
Noise ∠Vo = -80.38 deg

The magnitude of the noise would have been higher at DC because the Bode plot has a peak near the noise frequency.

d: Linear graph model of the filter

The linear graph model is shown in .

e: Filter transfer function \(H(s)\)

syms s
syms R L C
syms ZR ZL ZC
assume(R>=0);
assume(L>=0);
assume(C>=0);
imp.ZR = R;
imp.ZL = L*s;
imp.ZC = 1/(C*s);
Zo = 1/(1/ZC + 1/ZR);
HZ = simplify( ...
    Zo/(Zo+ZL) ...
)
H = simplify( ...
    subs(HZ,imp) ...
)

\(HZ =\frac{\textrm{ZC}\,\textrm{ZR}}{\textrm{ZC}\,\textrm{ZL}+\textrm{ZC}\,\textrm{ZR}+\textrm{ZL}\,\textrm{ZR}}\)

\(H =\frac{R}{C\,L\,R\,s^2 +L\,s+R}\)

f: Filter natural freq. and damping ratio

Use the denominator of the transfer function, like

[Hn,Hd] = numden(H);
co = coeffs(Hd,s);
a0 = co(1)/co(3);
a1 = co(2)/co(3);
wnf = sqrt(a0)
znf = simplify( ...
    a1/(2*wnf) ...
)
Kf = limit(H,s,0)

\(wnf =\sqrt{\frac{1}{C\,L}}\)

\(znf =\frac{\sqrt{L}}{2\,\sqrt{C}\,R}\)

\(Kf =1\)

g: Filter design

psol = solve([znf == 1,wnf == 1000,C == 1000e-6],[R,L,C]);
R_ = double(psol.R);
L_ = double(psol.L);
C_ = double(psol.C);
fprintf("R = %.4g Ω\nL = %.4g H\nC = %.4g F",R_,L_,C_)
R = 0.5 Ω
L = 0.001 H
C = 0.001 F

h: Transfer function \(V_o/A\)

Since the systems do not load each other, the combined transfer function \(V_o(s)/A(s)\) is simply \[\begin{aligned} \frac{V_o(s)}{A(s)} &= G(s) H(s). \end{aligned}\]

Substitute in parameter values and convert the symbolic transfer function for the filter \(H(s)\) into a Matlab tf object:

H_ = simplify(subs(H,psol))
H_ = sym_to_tf(H_,s) % using matlab-rico

\(H_ =\frac{1000000}^2 }\)

H_ =
 
         1e06
  -------------------
  s^2 + 2000 s + 1e06
 
Continuous-time transfer function.

Now we can find the combined transfer function

Vo_A = G*H_
Vo_A =
 
                       9e11
  -----------------------------------------------
  s^4 + 3200 s^3 + 1.24e07 s^2 + 1.92e10 s + 9e12
 
Continuous-time transfer function.

i: Bode plot for \(V_o/A\)

Now we can see the new Bode plot, shown in , with

figure;
bode(Vo_A);
grid on;

j: Noise improvement with filter

We should expect the noise output to be lower because, as shown in , the transfer function \(V_o/A\) has lower magnitude at the noise frequency than does the transfer function \(G\) (with its Bode plot shown in ).

k: Filtered noise output

GHnoise = freqresp(Vo_A,wnoise); % FRF
Vonoise_mag = abs(GHnoise)*anoise;
Vonoise_pha = 0 + angle(GHnoise) - 2*pi;
fprintf("Noise |Vo| = %.4g V\n",Vonoise_mag)
fprintf("Noise ∠Vo = %.4g deg",rad2deg(Vonoise_pha))
Noise |Vo| = 0.0271 V
Noise ∠Vo = -222.3 deg

l: Noise output comparison

The ratio of the filtered noise magnitude over the unfiltered noise magnitude is

fprintf("Noise ratio = %.4g", ...
    Vonoise_mag/Vinoise_mag ...
)
Noise ratio = 0.1063

This means the filtered noise magnitude is about \(1/10\) that of the unfiltered noise.

m: Augmenting the filter

We could further reduce the noise by decreasing the filter’s natural frequency, which we set at \(1000\) rad/s. The lower this frequency, the more aggressive will be the filter. The downside is that the dynamic response of the measurement will be slower (a smaller bandwidth yields slower responses).

Problem 12.7 (GYPSUM)

Respond to the following questions and imperatives with a sentence or two, equation, and/or a sketch.

  1. Comment on the stability and transient response characteristics of a system with eigenvalues \[\begin{aligned} -2, -5, -8 + j 3, -8 - j 3. \end{aligned}\]

  2. Consider an LTI system that, given input \(u_1\), outputs \(y_1\), and given input \(u_2\), outputs \(y_2\). If the input is \(u_3 = 5 u_1 - 6 u_2\), what is the output \(y_3\)?

  3. Consider a second-order system with natural frequency \(\omega_n = 2\) rad/s and damping ratio \(\zeta = 0.5\). What is the free response for initial condition \(y(0) = 1\)?

  4. Two thermal elements with impedances \(Z_1\) and \(Z_2\) have a temperature source \(T_S\) applied across them in series. What is the transfer function from \(T_S\) to the heat \(Q_2\) through \(Z_2\)?

  5. Draw a linear graph of a pump (pressure source) flowing water through a long pipe into the bottom of a tank, which has a valve at its bottom from which the water flows.


  1. Assume the input phase is zero.↩︎

  2. Be cautious to make the denominator have the proper standard form \(s^2 + 2 \zeta_H \omega_{n_H}\,s + \omega_{n_H}^2\).↩︎

Online Resources for Section 12.7

No online resources.