22-Mec-A3 System Analysis and Control · December 2014
Question 2 of 6: Bode Diagrams of Two Transfer Functions
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Notes on this paper
Paper format. National Exams, December 2014 —
07-Mec-A3 System Analysis and Control. Three hours, closed book
(Casio or Sharp approved calculator; semi-logarithmic graph paper is the only permitted aid).
Six questions are printed; the rubric states that any four questions constitute a complete
paper and that all questions are of equal value, so each carries 25 marks.
A two-page table of Laplace transform pairs is appended to the paper.
All six questions are solved here.
Reference texts. K. Ogata, Modern Control Engineering, 5th ed.
(Prentice Hall); N. S. Nise, Control Systems Engineering, 8th ed. (Wiley);
R. C. Dorf & R. H. Bishop, Modern Control Systems, 13th ed. (Pearson);
G. F. Franklin, J. D. Powell & A. Emami-Naeini, Feedback Control of Dynamic
Systems, 8th ed. (Pearson). Section references in the concept panels below follow
Ogata 5th ed. and Nise 8th ed.
Convention used throughout. Angles are quoted in
degrees for phasor results and in radians inside time functions; a lagging phase is written
as a negative angle. "Time constant" always means the reciprocal of the magnitude of the
real part of the governing closed-loop pole.
Question 2: Bode Diagrams of Two Transfer Functions (25 marks)
Given. Two open-loop transfer functions, each already in a factored form
with unity leading coefficients:
Part
$G(s)$
Poles
Zeros
System type
(a)
$1/[s^{2}(s+1)]$
$s=0$ (double), $s=-1$
none
type 2
(b)
$s/(s+1)^{2}$
$s=-1$ (double)
$s=0$
type 0 with a differentiator
Find. Magnitude (dB) and phase (degrees) versus $\log\omega$ for each
transfer function: the asymptote slopes, the corner frequencies, the exact values at the
corners, and the resulting shape of the two curves.
Approach. Put each factor in the time-constant (Bode) form
$(1+s/p)$ so that the low-frequency asymptote reads directly off the gain constant, add the
individual asymptotic contributions ($\pm20$ dB/decade per real factor, $\pm90^\circ$ per
real factor), then correct the asymptotes at each corner with the exact
$-3$ dB / $+3$ dB deviation of a first-order factor.
Part (a): $G(s)=1/[s^{2}(s+1)]$
Identify the Bode constant and the corner. The denominator factor
$(s+1)$ is already in time-constant form, so the Bode gain is $K_{B}=1$ and the single
corner frequency is $\omega_{c}=1$ rad/s. The double pole at the origin contributes
$1/(j\omega)^{2}$ at every frequency.
Build the magnitude asymptotes. Below the corner only the double
integrator acts, giving a slope of $-40$ dB/decade passing through $0$ dB at
$\omega=1$ rad/s. Above the corner the pole at $-1$ adds a further $-20$ dB/decade, so the
asymptote steepens to $-60$ dB/decade. Sample values on the low-frequency asymptote:
$$|G(j0.1)| = \frac{1}{0.1^{2}\sqrt{1+0.1^{2}}} = 99.5 \;\Rightarrow\; 39.96\ \text{dB}.$$
At the corner the exact magnitude falls $3$ dB below the asymptote's $0$ dB, so
$$\boxed{\;|G(j1)| = \tfrac{1}{\sqrt2}= -3.01\ \text{dB},\qquad
\text{slope } -40 \to -60\ \text{dB/decade at }\omega=1\ \text{rad/s}. \;}$$
Locate the gain crossover. Setting $|G(j\omega)|=1$ requires
$\omega^{2}\sqrt{\omega^{2}+1}=1$, which solves numerically to
$\omega_{gc}=0.8688$ rad/s — slightly below the corner, as the $-3$ dB correction
implies.
Build the phase curve. The double integrator contributes a constant
$-180^\circ$ and the real pole a further $-\tan^{-1}\omega$. Hence
$\angle G = -180^{\circ}-\tan^{-1}\omega$, which starts at $-180^\circ$, passes through
$-225^{\circ}$ at the corner $\omega=1$, and approaches $-270^{\circ}$ at high frequency.
The phase is below $-180^\circ$ at every frequency, so the phase margin is
negative. Evaluated at the gain crossover found in step 3 — not at the corner —
$\angle G(j0.8688) = -180^{\circ}-\tan^{-1}(0.8688) = -220.99^{\circ}$, so
$\mathrm{PM} = 180^{\circ}+\angle G = -40.99^{\circ}$. This plant cannot be stabilised by
proportional gain alone and needs phase lead.
Bode diagram of $1/[s^{2}(s+1)]$. Magnitude
begins at $-40$ dB/decade and breaks to $-60$ dB/decade at $\omega=1$ rad/s, with the usual
$-3$ dB deviation at the corner; phase runs from $-180^{\circ}$ to $-270^{\circ}$, never
above $-180^{\circ}$.
Part (b): $G(s)=s/(s+1)^{2}$
The printed denominator $(s+1)(s+1)$ is a repeated first-order factor, so
$G(s)=s/(s+1)^{2}$: a single differentiating zero at the origin against a double pole at
$-1$ rad/s.
Set the low-frequency behaviour. For $\omega\ll1$ the denominator is
approximately unity and $G\approx j\omega$, so the magnitude asymptote rises at
$+20$ dB/decade through $0$ dB at $\omega=1$ rad/s, and the phase sits at $+90^{\circ}$.
At $\omega=0.1$ the asymptote reads $-20$ dB.
Set the high-frequency behaviour. For $\omega\gg1$,
$G\approx j\omega/(j\omega)^{2}=1/(j\omega)$, so the magnitude falls at
$-20$ dB/decade and the phase settles at $-90^{\circ}$. The net slope change at the corner
is $-40$ dB/decade, because both repeated poles break at the same frequency.
Evaluate the peak. The magnitude is
$|G(j\omega)| = \omega/(1+\omega^{2})$, whose maximum is found from
$d/d\omega\,[\omega/(1+\omega^{2})]=0 \Rightarrow \omega=1$ rad/s. There
$$\boxed{\;|G(j1)| = \tfrac{1}{2} = -6.02\ \text{dB},\qquad
\angle G(j1)=90^{\circ}-2\tan^{-1}(1)=0^{\circ}. \;}$$
The exact curve therefore lies $6$ dB below the intersection of the two asymptotes, which is
the standard correction for a repeated first-order pole (twice the usual 3 dB) and
is the single most important number to place on the sketch.
Build the phase curve. $\angle G = 90^{\circ}-2\tan^{-1}\omega$,
which decreases monotonically from $+90^{\circ}$ through $0^{\circ}$ at the corner to
$-90^{\circ}$. Because the phase never reaches $-180^{\circ}$, this plant has an infinite
gain margin — the exact opposite of part (a), and a useful contrast to draw in the
answer.
Bode diagram of $s/(s+1)^{2}$. The
band-pass magnitude peaks at $-6.02$ dB at $\omega=1$ rad/s (a repeated pole gives twice the
usual corner correction); phase falls from $+90^{\circ}$ to $-90^{\circ}$ and never reaches
$-180^{\circ}$.