CHAPTER 25 / FREQUENCY DOMAIN

Drawing Bode plots by hand and computing margins

Split the loop into simple factors, draw the straight-line asymptotes of each factor, add them, and read the gain margin and phase margin from the result.

THE SHORT ANSWER

Drawing a Bode plot by hand means writing the loop in Bode form, splitting it into gain, integrator, first-order and quadratic factors, and adding each factor's straight-line asymptotes in decibels and its phase in degrees. The magnitude slope changes at each corner frequency: −20 dB/dec for a pole, +20 dB/dec for a zero, −40 dB/dec for a quadratic pole pair. Gain and phase margins are then read at the two crossovers.

20log⁡10∣L(jω)∣=∑i20log⁡10∣Li(jω)∣,∠L(jω)=∑i∠Li(jω)20\log_{10}|L(j\omega)|=\sum_i20\log_{10}|L_i(j\omega)|,\qquad \angle L(j\omega)=\sum_i\angle L_i(j\omega)
WHAT YOU WILL LEARN

Draw the asymptotic Bode plot of a transfer function from its factors and compute its gain crossover frequency, phase crossover frequency, gain margin and phase margin.

The Bode chapter defined the magnitude and phase plots and the 0 dB line. Now we construct those plots by hand for any rational L(s), so that the crossover frequencies and the margins become numbers you can compute on paper.

Review the previous lesson: Bode plots: gain and phase
START WITH AN EXAMPLE

First, picture it

A transfer function such as 1/(s(s+1)(s+2)) looks as if it needs a computer to plot. It does not. A Bode magnitude is in decibels, and a decibel value is a logarithm, so a product of factors becomes a sum of curves. The phase angles of a product add as well. Each factor is one of four kinds: a gain, an integrator, a first-order term or a quadratic term. The magnitude curve of each is close to two straight lines that meet at a corner frequency. So you draw a few straight lines, add up their slopes from left to right, and have the whole magnitude plot to within about 3 dB. The phase plot is one arctangent per factor, added the same way.

THE IDEA

Build the idea step by step

First put L(s) in Bode form. Write every first-order factor as (1 + sT) and every quadratic factor as 1 + 2ζ(s/ωn) + (s/ωn)², so that each factor equals 1 at low frequency and all the leftover constants collect into one number, the Bode gain K_B. Next list the corner frequencies in increasing order: 1/T for each first-order factor and ωn for each quadratic factor. Start the magnitude asymptote at the far left, with the slope set by the integrators, −20 dB/dec each, and the height set by K_B. Move to the right and change the slope at every corner: −20 dB/dec for a pole, +20 dB/dec for a zero, −40 dB/dec for a quadratic pole pair. Compute the phase at a handful of frequencies by adding the angles of the factors. Finally read the two crossovers. The frequency where the magnitude crosses 0 dB gives the phase margin, and the frequency where the phase crosses −180° gives the gain margin.

Read the equations

L(s)=KB (1+sTa)⋯sN (1+sT1)⋯[1+2ζsωn+s2ωn2]⋯L(s)=\frac{K_B\,(1+sT_a)\cdots}{s^N\,(1+sT_1)\cdots\left[1+2\zeta\dfrac{s}{\omega_n}+\dfrac{s^2}{\omega_n^2}\right]\cdots}

Bode form, also called time-constant form. Rewrite every (s + a) as a(1 + s/a) and collect the constants into the Bode gain K_B. Each bracket now equals 1 at ω = 0.

20log⁡10∣L(jω)∣=∑i20log⁡10∣Li(jω)∣,∠L(jω)=∑i∠Li(jω)20\log_{10}|L(j\omega)|=\sum_i20\log_{10}|L_i(j\omega)|,\qquad \angle L(j\omega)=\sum_i\angle L_i(j\omega)

Additivity. Magnitudes in dB add and phases add, so the Bode plot of L is the sum of the plots of its factors. A factor in the numerator contributes the negative of what the same factor contributes in the denominator.

KB:20log⁡10KB dB,0∘(KB>0)K_B:\qquad 20\log_{10}K_B\ \mathrm{dB},\qquad 0^\circ\quad(K_B>0)

Gain factor. A horizontal line at 20 log₁₀K_B and no phase. Changing the gain moves the whole magnitude curve up or down and leaves the phase curve where it is.

1(jω)N:−20Nlog⁡10ω dB,−90∘N\frac{1}{(j\omega)^N}:\qquad -20N\log_{10}\omega\ \mathrm{dB},\qquad -90^\circ N

Integrator factor. A straight line of slope −20N dB/dec that passes through 0 dB at ω = 1 rad/s, with a constant phase of −90°N. For a differentiator (jω)^N both signs flip: +20N dB/dec and +90°N.

∣11+jωT∣dB=−10log⁡10(1+ω2T2)≈{0,ω<1/T−20log⁡10(ωT),ω>1/T\left|\frac{1}{1+j\omega T}\right|_{\mathrm{dB}}=-10\log_{10}(1+\omega^2T^2)\approx\begin{cases}0,&\omega<1/T\\-20\log_{10}(\omega T),&\omega>1/T\end{cases}

First-order pole, magnitude. The asymptote is 0 dB up to the corner frequency 1/T and falls at −20 dB/dec beyond it. The true curve is 3 dB below the asymptote at the corner, which is the largest error, and 1 dB below it one octave to either side. A first-order zero (1 + jωT) is the mirror image and rises at +20 dB/dec.

∠11+jωT=−tan⁡−1(ωT):0∘ ⟶ −45∘ (ω=1T) ⟶ −90∘\angle\frac{1}{1+j\omega T}=-\tan^{-1}(\omega T):\qquad 0^\circ\ \longrightarrow\ -45^\circ\ \left(\omega=\tfrac1T\right)\ \longrightarrow\ -90^\circ

First-order pole, phase. The phase goes from 0° to −90° and is exactly −45° at the corner. Nearly all of the change happens between 0.1/T and 10/T. A first-order zero gives the same angles with a plus sign.

G2(jω)=11+2ζ (jω/ωn)+(jω/ωn)2,∣G2∣dB≈{0,ω<ωn−40log⁡10(ω/ωn),ω>ωn∣G2(jωn)∣=12ζ,∠G2(jωn)=−90∘\begin{gathered}G_2(j\omega)=\frac{1}{1+2\zeta\,(j\omega/\omega_n)+(j\omega/\omega_n)^2},\qquad |G_2|_{\mathrm{dB}}\approx\begin{cases}0,&\omega<\omega_n\\-40\log_{10}(\omega/\omega_n),&\omega>\omega_n\end{cases}\\[4pt]|G_2(j\omega_n)|=\frac{1}{2\zeta},\qquad \angle G_2(j\omega_n)=-90^\circ\end{gathered}

Quadratic factor. The corner frequency is ωn, and beyond it the asymptote falls at −40 dB/dec. At the corner the true magnitude is 1/(2ζ) and the phase is −90°; the phase runs from 0° to −180° in total.

ωr=ωn1−2ζ2,Mr=12ζ1−ζ2(0<ζ<0.707)\omega_r=\omega_n\sqrt{1-2\zeta^2},\qquad M_r=\frac{1}{2\zeta\sqrt{1-\zeta^2}}\qquad(0<\zeta<0.707)

Resonant peak. For ζ below 0.707 the magnitude of the quadratic factor rises to a peak M_r at the resonant frequency ωr, slightly below ωn. For ζ of 0.707 or more there is no peak.

∣L(jωgc)∣=1,PM=180∘+∠L(jωgc)|L(j\omega_{\mathrm{gc}})|=1,\qquad \mathrm{PM}=180^\circ+\angle L(j\omega_{\mathrm{gc}})

Gain crossover and phase margin. Find the frequency where the magnitude crosses 0 dB, compute the phase there, and add 180°. The result is how much more phase lag the loop can take at that frequency before it reaches −180°.

∠L(jωpc)=−180∘,GM=1∣L(jωpc)∣,GMdB=−20log⁡10∣L(jωpc)∣\angle L(j\omega_{\mathrm{pc}})=-180^\circ,\qquad \mathrm{GM}=\frac{1}{|L(j\omega_{\mathrm{pc}})|},\qquad \mathrm{GM_{dB}}=-20\log_{10}|L(j\omega_{\mathrm{pc}})|

Phase crossover and gain margin. Find the frequency where the phase crosses −180° and compute the magnitude there. The gain margin is the factor by which the loop gain can grow before that magnitude reaches 1.

d ∣L∣dBdlog⁡10ω≈−20n dB/dec ⟹ ∠L(jω)≈−90∘n\frac{d\,|L|_{\mathrm{dB}}}{d\log_{10}\omega}\approx-20n\ \mathrm{dB/dec}\ \Longrightarrow\ \angle L(j\omega)\approx-90^\circ n

Slope and phase in a minimum-phase system. Where the magnitude slope holds at −20n dB/dec for about a decade, the phase is close to −90°n. Design rule of thumb: cross 0 dB at −20 dB/dec to get a comfortable phase margin.

e−jωTd:0 dB,−ωTd rad=−57.3∘×ωTde^{-j\omega T_d}:\qquad 0\ \mathrm{dB},\qquad -\omega T_d\ \mathrm{rad}=-57.3^\circ\times\omega T_d

Time delay. A pure delay leaves the magnitude plot unchanged and subtracts a phase that grows in proportion to frequency. It lowers the phase margin by 57.3° × ωgc × T_d.

What the symbols mean

L(s)L(s)
open-loop transfer function
ω\omega
angular frequency, rad/s
KBK_B
Bode gain: the constant left in front after every factor is normalized to 1 at low frequency; its unit depends on N
NN
number of integrators in L(s), the system type
TT
time constant of a first-order factor, s; its corner frequency is 1/T in rad/s
ζ\zeta
damping ratio of a quadratic factor, dimensionless
ωn\omega_n
undamped natural frequency and corner frequency of a quadratic factor, rad/s
ωr, Mr\omega_r,\,M_r
resonant frequency in rad/s and resonant peak as a magnitude ratio
ωgc\omega_{\mathrm{gc}}
gain crossover frequency, where |L| = 1 (0 dB), rad/s
ωpc\omega_{\mathrm{pc}}
phase crossover frequency, where the phase is −180°, rad/s
PM\mathrm{PM}
phase margin, degrees
GM\mathrm{GM}
gain margin, as a ratio or in dB
TdT_d
time delay, s

The margin formulas are used here for a loop that is open-loop stable and minimum phase, whose magnitude crosses 0 dB once and whose phase crosses −180° once. For such a loop, positive gain and phase margins mean the closed loop is stable; in any other case use the Nyquist criterion of the next chapter.

WORKED EXAMPLE

Bode plot and margins of K/(s(s+1)(s+2))

  1. Bode form. Take the 2 out of (s + 2). For K = 1 the Bode gain is 0.5, which is −6.02 dB.

    L(s)=Ks(s+1)(s+2)=K/2s (1+s) (1+s/2),KB=0.5L(s)=\frac{K}{s(s+1)(s+2)}=\frac{K/2}{s\,(1+s)\,(1+s/2)},\qquad K_B=0.5
  2. Factors and slopes. There is a gain, one integrator, a pole with corner 1 rad/s and a pole with corner 2 rad/s. The slope starts at −20 dB/dec and steepens by 20 dB/dec at each corner.

    −20⏟ω<1 ⟶ −40⏟1<ω<2 ⟶ −60⏟ω>2dB/dec\underbrace{-20}_{\omega<1}\ \longrightarrow\ \underbrace{-40}_{1<\omega<2}\ \longrightarrow\ \underbrace{-60}_{\omega>2}\quad\mathrm{dB/dec}
  3. Heights of the asymptote. The low-frequency line is 0.5/ω: it is +13.98 dB at 0.1 rad/s, crosses 0 dB at 0.5 rad/s and reaches −6.02 dB at the first corner. From there it falls 40 dB/dec to the second corner. The true curve at 1 rad/s is at −10.0 dB, 3.98 dB below the asymptote.

    ω=1: −6.02 dB,ω=2: −6.02−40log⁡102=−18.06 dB\omega=1:\ -6.02\ \mathrm{dB},\qquad \omega=2:\ -6.02-40\log_{10}2=-18.06\ \mathrm{dB}
  4. Phase. Add the integrator's −90° and one arctangent for each pole, and evaluate at a few frequencies. The phase passes −180° somewhere between 1 and 2 rad/s.

    ∠L(jω)=−90∘−tan⁡−1ω−tan⁡−1ω2:ω0.10.51210∠L−98.6∘−130.6∘−161.6∘−198.4∘−253.0∘\angle L(j\omega)=-90^\circ-\tan^{-1}\omega-\tan^{-1}\frac{\omega}{2}:\quad \begin{array}{c|ccccc}\omega&0.1&0.5&1&2&10\\\hline\angle L&-98.6^\circ&-130.6^\circ&-161.6^\circ&-198.4^\circ&-253.0^\circ\end{array}
  5. Phase crossover. The two arctangents must add to 90°, which happens when the product of their arguments is 1.

    tan⁡−1ω+tan⁡−1ω2=90∘  ⟺  ω⋅ω2=1 ⇒ ωpc=2≈1.414 rad/s\tan^{-1}\omega+\tan^{-1}\frac{\omega}{2}=90^\circ\iff\omega\cdot\frac{\omega}{2}=1\ \Rightarrow\ \omega_{\mathrm{pc}}=\sqrt2\approx1.414\ \mathrm{rad/s}
  6. Gain margin. Evaluate the magnitude at the phase crossover. For K = 1 the gain margin is 6, which is 15.56 dB.

    ∣L(j2)∣=K2⋅3⋅6=K6,GM=6K|L(j\sqrt2)|=\frac{K}{\sqrt2\cdot\sqrt3\cdot\sqrt6}=\frac{K}{6},\qquad \mathrm{GM}=\frac{6}{K}
  7. Gain crossover for K = 1. The asymptote crosses 0 dB at 0.5 rad/s, but the true curve is 1.23 dB lower there, so the true crossover is a little to the left. Solve |L| = 1 by trial: ω = 0.45 gives |L| = 0.989 and ω = 0.44 gives 1.016.

    ω2(ω2+1)(ω2+4)=1 ⇒ ωgc≈0.446 rad/s\omega^2(\omega^2+1)(\omega^2+4)=1\ \Rightarrow\ \omega_{\mathrm{gc}}\approx0.446\ \mathrm{rad/s}
  8. Phase margin for K = 1. Evaluate the phase at the gain crossover and add 180°.

    PM=180∘−90∘−tan⁡−1(0.446)−tan⁡−1(0.223)=90∘−24.0∘−12.6∘=53.4∘\mathrm{PM}=180^\circ-90^\circ-\tan^{-1}(0.446)-\tan^{-1}(0.223)=90^\circ-24.0^\circ-12.6^\circ=53.4^\circ
  9. Raise the gain to K = 6. The magnitude curve moves up by 20 log₁₀6 = 15.56 dB and the phase curve does not move, so the 0 dB crossing lands exactly on the phase crossover and both margins are zero. This is the root-locus result seen from the frequency side: at K = 6 the closed-loop poles are at ±j√2.

    K=6:ωgc=ωpc=2,GM=0 dB,PM=0∘K=6:\quad \omega_{\mathrm{gc}}=\omega_{\mathrm{pc}}=\sqrt2,\qquad \mathrm{GM}=0\ \mathrm{dB},\qquad \mathrm{PM}=0^\circ
REFERENCE

Reference tables

Bode asymptotes and phase of each factor
FactorMagnitude asymptoteCorner frequencyPhaseTrue magnitude at the corner
KB>0K_B>0flat at 20 log₁₀K_B dB—0°—
1(jω)N\dfrac{1}{(j\omega)^N}−20N dB/dec, through 0 dB at ω = 1—−90°N—
(jω)N(j\omega)^N+20N dB/dec, through 0 dB at ω = 1—+90°N—
11+jωT\dfrac{1}{1+j\omega T}0 dB, then −20 dB/dec1/T0° → −45° → −90°−3.01 dB (−0.97 dB one octave away)
1+jωT1+j\omega T0 dB, then +20 dB/dec1/T0° → +45° → +90°+3.01 dB
11+2ζjωωn+(jωωn)2\dfrac{1}{1+2\zeta\frac{j\omega}{\omega_n}+\left(\frac{j\omega}{\omega_n}\right)^2}0 dB, then −40 dB/decωn0° → −90° → −180°−20 log₁₀(2ζ) dB
1+2ζjωωn+(jωωn)21+2\zeta\frac{j\omega}{\omega_n}+\left(\frac{j\omega}{\omega_n}\right)^20 dB, then +40 dB/decωn0° → +90° → +180°+20 log₁₀(2ζ) dB
e−jωTde^{-j\omega T_d}flat at 0 dB—−57.3° × ωT_d, without limit—

Zeros are assumed to lie in the left half-plane (T > 0, ζ > 0). Add the magnitude asymptotes in dB and the phases in degrees to get the plot of the whole loop.

Quadratic pole factor near its corner frequency
ζ\zeta|G₂(jωn)| (dB)Resonant peak Mr (dB)ωr/ωn\omega_r/\omega_n
0.113.9814.020.990
0.27.968.140.959
0.34.444.850.906
0.50.001.250.707
0.707 (1/√2)−3.01no peak—
1.0−6.02no peak—

The asymptote is 0 dB at ωn, so the second column is the correction to draw there. Below ζ = 0.707 the peak sits at ωr = ωn√(1 − 2ζ²) and is slightly higher than the value at ωn.

More detail and model assumptions
A CLOSER LOOK

Why decibels make the plot additive

Write the loop as a product, L(jω) = L₁(jω)L₂(jω)⋯. The magnitude of a product is the product of the magnitudes, and taking 20 log₁₀ turns that product into a sum. The angle of a product is the sum of the angles without any extra step. So the Bode plot of L is the Bode plots of its factors added point by point. A factor in the numerator is the mirror image of the same factor in the denominator, with both magnitude in dB and phase changing sign. That is why four factor types cover every rational transfer function.

The logarithmic frequency axis is what makes the pieces straight. For ω much larger than 1/T, the magnitude of 1/(1 + jωT) is close to 1/(ωT), which is −20 log₁₀(ωT) dB. Against log ω this is a straight line that falls 20 dB for every tenfold increase in frequency, or 6 dB per octave. For ω much smaller than 1/T the magnitude is close to 1, which is 0 dB. The two lines meet at ω = 1/T, and that is why 1/T is called the corner frequency.

Bode form puts all the low-frequency information in one place. Every normalized factor contributes 0 dB well below its corner, so the low-frequency asymptote is K_B/(jω)^N alone. This line, or its extension, passes through 20 log₁₀K_B at ω = 1 rad/s. For a type-1 loop it crosses 0 dB at ω = K_B, and K_B is the velocity error constant K_v from the chapter on system type.

A CLOSER LOOK

How far the asymptotes are from the true curve

For a first-order pole the true magnitude is 3.01 dB below the asymptote at the corner, 0.97 dB below it one octave to either side, and only 0.04 dB below it a decade away. To correct a sketch, mark 3 dB at the corner and 1 dB at half and at double the corner frequency, then draw a smooth curve through the marks. Errors from nearby corners add. In the worked example the true curve at ω = 1 rad/s is 3.98 dB below the asymptote: 3.01 dB from the corner at 1 rad/s and 0.97 dB from the corner at 2 rad/s.

The phase of a first-order pole is −tan⁻¹(ωT): −5.7° at 0.1/T, −26.6° at 0.5/T, −45° at the corner, −63.4° at 2/T and −84.3° at 10/T. A common straight-line sketch holds 0° up to 0.1/T, falls at −45° per decade through the corner, and holds −90° beyond 10/T; it is never more than about 6° off. For a margin, compute the phase from the arctangents instead of reading it from that sketch, because the calculation takes one line and the margin depends on a few degrees.

A quadratic factor needs more care, because its error at the corner depends on ζ. At ω = ωn the true magnitude is 1/(2ζ) times the asymptote: +14 dB for ζ = 0.1, 0 dB for ζ = 0.5 and −6 dB for ζ = 1. The phase is −90° at ωn for every ζ, but the smaller ζ is, the more abruptly the phase swings from 0° to −180° around the corner. For a lightly damped factor, always compute the resonant peak and mark it before drawing the curve.

A CLOSER LOOK

Minimum phase, slope and margin

A transfer function is minimum phase when all its poles and zeros are in the left half-plane and it contains no time delay. For such a system the magnitude curve fixes the phase curve: Bode showed that the phase at a frequency is set mainly by the slope of the magnitude near that frequency. As a working approximation, if the slope stays at −20n dB/dec for about a decade around ω, the phase there is close to −90°n.

This gives the design rule of thumb: make the magnitude cross 0 dB with a slope of −20 dB/dec, and keep that slope over a reasonable band on both sides of the crossover. The phase at crossover is then not far below −90° and the phase margin is comfortable. A crossing at −40 dB/dec puts the phase near −180° and leaves little or no margin. The worked example shows both cases. At K = 1 the crossover at 0.446 rad/s is on the −20 dB/dec segment, and the corners at 1 and 2 rad/s, less than a decade away, pull the margin down from 90° to 53°. At K = 6 the crossover at √2 rad/s lies between the corners, where the slope is already −40 dB/dec, and the margin is zero.

Time delay and right-half-plane zeros break the link between slope and phase. The factor e^(−sT_d) has magnitude 1 at every frequency, so it is invisible on the magnitude plot, yet it subtracts ωT_d radians of phase, a lag that grows without limit as ω increases. A right-half-plane zero (1 − sT) has the same magnitude as (1 + sT) but contributes phase lag instead of lead. With either of these in the loop, compute the phase explicitly; for a delay, the largest tolerable value is the phase margin in radians divided by the gain crossover frequency.

TRY IT

Connect the example to the lab

The explorer in the Bode chapter (/theory/bode) plots K/(s(1 + s/p)). At its starting setting K = 10, p = 10 it reports a gain crossover of 7.86 rad/s and a phase margin of 52°. Check this by hand: ω²(1 + ω²/100) = 100 gives ω = 7.86 rad/s, and 90° − tan⁻¹(0.786) = 51.8°. Then move K and watch the magnitude curve slide up and down while the phase curve stays where it is, exactly as the gain-factor rule says. The motor lab linked below plots the Bode magnitude and phase of the motor loop and prints its gain crossover, phase margin and gain margin, so you can compare your reading of the curves with computed numbers. That loop is sampled, so its phase falls faster at high frequency than a continuous-time sketch predicts, in the same way as the time-delay factor at the end of the equation list.

Open the motor loop's Bode plot and margins
YOUR EXPERIMENT

Make a prediction before moving a control.

  1. PREDICTThe gain-factor rule says that doubling the whole loop gain raises the magnitude curve by 6 dB at every frequency and leaves the phase alone. In the motor lab you will double only the proportional gain Kp of a PID controller. Predict: will the magnitude rise by 6 dB everywhere, and in which frequency range will it hardly move?
  2. CHANGE ONE THINGOpen the lesson, open the Bode view of the frequency workbench and note the gain crossover, phase margin and gain margin. Then double Kp and change nothing else.
  3. OBSERVECompare the magnitude curve before and after at the low-frequency end, around the crossover and at the high-frequency end. Read the new gain crossover and both margins.
  4. EXPLAINWrite the PI part of the controller in Bode form, Ki(1 + sKp/Ki)/s. Its low-frequency asymptote depends on Ki only, and its zero has the corner frequency Ki/Kp, about 1.5 rad/s for the preset Kp = 0.4 and Ki = 0.6. Use this to explain where the curve moved, where it did not, and why the phase curve changed even though only a gain was touched.
BE CAREFUL

Convert to Bode form before you read the gain. In L(s) = 1/(s(s+1)(s+2)) the low-frequency gain is 1/2, not 1, because (s + 2) = 2(1 + s/2) hides a factor of 2. Skipping this step shifts the whole magnitude plot by 6 dB and makes both margins wrong.

Check your understanding+

L(s) = K/(s(s+1)(s+2)) has a gain margin of 15.6 dB at K = 1. What is the gain margin at K = 2, and does the phase crossover frequency move?

GM = 6/2 = 3, which is 9.5 dB. Doubling K raises the magnitude curve by 6 dB and leaves the phase curve untouched, so the phase crossover stays at √2 rad/s.
PRACTICE PROBLEMS

Solve these before opening the solutions

Work each problem on paper first. Open the answer to check yours, then the full solution if you need the steps, and mark how it went.

  1. Write L(s) in Bode form. List the corner frequencies and the slope of the magnitude asymptote in each band, then find the height of the asymptote at 1 rad/s and the frequency where the asymptote crosses 0 dB.

    L(s)=100(s+5)s(s+50)L(s)=\frac{100(s+5)}{s(s+50)}
    Show answer

    AnswerK_B = 10. A zero corner at 5 rad/s and a pole corner at 50 rad/s. Slopes: −20 dB/dec below 5 rad/s, 0 between 5 and 50 rad/s, −20 dB/dec above 50 rad/s. The asymptote is at 20 dB at 1 rad/s, flat at 6.02 dB between the corners, and crosses 0 dB at 100 rad/s.

    L(s)=10 (1+s/5)s (1+s/50),ωgc≈100 rad/sL(s)=\frac{10\,(1+s/5)}{s\,(1+s/50)},\qquad \omega_{\mathrm{gc}}\approx100\ \mathrm{rad/s}
    Show full solution

    Solution

    1. Normalize each factor and collect the constants.

      100⋅5 (1+s/5)s⋅50 (1+s/50)=10 (1+s/5)s (1+s/50)\frac{100\cdot5\,(1+s/5)}{s\cdot50\,(1+s/50)}=\frac{10\,(1+s/5)}{s\,(1+s/50)}
    2. Slopes. The integrator gives −20 dB/dec. The zero at 5 rad/s adds +20, which makes the asymptote flat. The pole at 50 rad/s adds −20.

      −20⏟ω<5 ⟶ 0⏟5<ω<50 ⟶ −20⏟ω>50dB/dec\underbrace{-20}_{\omega<5}\ \longrightarrow\ \underbrace{0}_{5<\omega<50}\ \longrightarrow\ \underbrace{-20}_{\omega>50}\quad\mathrm{dB/dec}
    3. Heights. The low-frequency line 10/ω is at 20 dB at 1 rad/s and has fallen to 6.02 dB at 5 rad/s, where it becomes flat.

      20log⁡1010=20 dB,20−20log⁡105=6.02 dB20\log_{10}10=20\ \mathrm{dB},\qquad 20-20\log_{10}5=6.02\ \mathrm{dB}
    4. Crossover. Beyond 50 rad/s the asymptote must fall 6.02 dB at −20 dB/dec, which takes a factor of 2 in frequency. The exact magnitude at 100 rad/s is 0.896, or −0.96 dB, so the asymptote is within 1 dB.

      50×106.02/20=100 rad/s,∣L(j100)∣=100 ∣5+j100∣100 ∣50+j100∣=0.89650\times10^{6.02/20}=100\ \mathrm{rad/s},\qquad |L(j100)|=\frac{100\,|5+j100|}{100\,|50+j100|}=0.896
  2. Write G(s) in Bode form. Compute the exact magnitude in dB and the phase at 4, 8 and 40 rad/s, and compare each magnitude with the asymptote.

    G(s)=20s+4G(s)=\frac{20}{s+4}
    Show answer

    AnswerK_B = 5 (13.98 dB) with a corner at 4 rad/s. Exact values: 10.97 dB and −45° at 4 rad/s; 6.99 dB and −63.4° at 8 rad/s; −6.06 dB and −84.3° at 40 rad/s. The asymptote gives 13.98, 7.96 and −6.02 dB, so the errors are −3.01, −0.97 and −0.04 dB.

    G(s)=51+s/4:10.97, 6.99, −6.06 dB;−45∘, −63.4∘, −84.3∘G(s)=\frac{5}{1+s/4}:\qquad 10.97,\ 6.99,\ -6.06\ \mathrm{dB};\qquad -45^\circ,\ -63.4^\circ,\ -84.3^\circ
    Show full solution

    Solution

    1. Bode form. The gain is 20/4 = 5 and the corner is at 4 rad/s.

      G(s)=51+s/4,20log⁡105=13.98 dBG(s)=\frac{5}{1+s/4},\qquad 20\log_{10}5=13.98\ \mathrm{dB}
    2. At the corner, 4 rad/s. The asymptote is still at 13.98 dB; the exact curve is 3.01 dB lower and the phase is −45°.

      13.98−10log⁡10(1+12)=10.97 dB,−tan⁡−11=−45∘13.98-10\log_{10}(1+1^2)=10.97\ \mathrm{dB},\qquad -\tan^{-1}1=-45^\circ
    3. One octave above, 8 rad/s. The asymptote has fallen 6.02 dB to 7.96 dB.

      13.98−10log⁡10(1+22)=6.99 dB,−tan⁡−12=−63.4∘13.98-10\log_{10}(1+2^2)=6.99\ \mathrm{dB},\qquad -\tan^{-1}2=-63.4^\circ
    4. One decade above, 40 rad/s. The asymptote has fallen 20 dB to −6.02 dB and the exact curve is almost on it.

      13.98−10log⁡10(1+102)=−6.06 dB,−tan⁡−110=−84.3∘13.98-10\log_{10}(1+10^2)=-6.06\ \mathrm{dB},\qquad -\tan^{-1}10=-84.3^\circ
  3. Find ωn and ζ. Then find the magnitude and phase at the corner frequency, the resonant frequency and the resonant peak, and compare the exact magnitude at 60 rad/s with the asymptote.

    G(s)=36s2+3.6s+36G(s)=\frac{36}{s^2+3.6s+36}
    Show answer

    Answerωn = 6 rad/s and ζ = 0.3. At the corner |G| = 1.67 (4.44 dB) with phase −90°. The resonant frequency is 5.43 rad/s and the peak is M_r = 1.75 (4.85 dB). At 60 rad/s the asymptote gives −40 dB and the exact value is −39.93 dB.

    ωn=6,ζ=0.3,ωr=5.43 rad/s,Mr=1.75 (4.85 dB)\omega_n=6,\quad \zeta=0.3,\quad \omega_r=5.43\ \mathrm{rad/s},\quad M_r=1.75\ (4.85\ \mathrm{dB})
    Show full solution

    Solution

    1. Match the denominator with the standard form s² + (2ζωn)s + ωn². The gain at ω = 0 is 36/36 = 1, so the low-frequency asymptote is 0 dB.

      ωn2=36,2ζωn=3.6 ⇒ ωn=6 rad/s,ζ=0.3\omega_n^2=36,\quad 2\zeta\omega_n=3.6\ \Rightarrow\ \omega_n=6\ \mathrm{rad/s},\quad \zeta=0.3
    2. At the corner the real part of the denominator vanishes, so G is purely imaginary.

      G(j6)=36j 21.6=−j1.67 ⇒ ∣G∣=12ζ=1.67 (4.44 dB),∠G=−90∘G(j6)=\frac{36}{j\,21.6}=-j1.67\ \Rightarrow\ |G|=\frac{1}{2\zeta}=1.67\ (4.44\ \mathrm{dB}),\quad \angle G=-90^\circ
    3. Resonance. Since ζ is below 0.707 there is a peak, slightly below the corner frequency and slightly higher than the corner value.

      ωr=61−2(0.3)2=5.43 rad/s,Mr=12(0.3)1−0.32=1.75\omega_r=6\sqrt{1-2(0.3)^2}=5.43\ \mathrm{rad/s},\qquad M_r=\frac{1}{2(0.3)\sqrt{1-0.3^2}}=1.75
    4. One decade above the corner the asymptote has fallen 40 dB. The exact magnitude differs by less than 0.1 dB, and the phase is −176.5°.

      ∣G(j60)∣=36∣36−3600+j216∣=0.0101 ⇒ −39.93 dB|G(j60)|=\frac{36}{|36-3600+j216|}=0.0101\ \Rightarrow\ -39.93\ \mathrm{dB}
  4. (a) For K = 20 find the gain margin and the phase margin. (b) Find the K that gives a phase margin of 45°.

    L(s)=Ks (1+0.1s) (1+0.02s)L(s)=\frac{K}{s\,(1+0.1s)\,(1+0.02s)}
    Show answer

    Answer(a) The phase crossover is at 22.4 rad/s, where |L| = K/60, so GM = 3 (9.54 dB). The gain crossover is at about 12.3 rad/s and PM ≈ 25.4°. (b) The phase is −135° at 7.42 rad/s, and placing the gain crossover there needs K ≈ 9.33.

    GM=3 (9.54 dB),PM≈25.4∘;PM=45∘ ⇒ K≈9.33\mathrm{GM}=3\ (9.54\ \mathrm{dB}),\quad \mathrm{PM}\approx25.4^\circ;\qquad \mathrm{PM}=45^\circ\ \Rightarrow\ K\approx9.33
    Show full solution

    Solution

    1. Phase crossover and gain margin. The two arctangents add to 90° when the product of their arguments is 1.

      0.1ω⋅0.02ω=1 ⇒ ωpc=500=22.4;∣L∣=K22.461.2=K60 ⇒ GM=6020=30.1\omega\cdot0.02\omega=1\ \Rightarrow\ \omega_{\mathrm{pc}}=\sqrt{500}=22.4;\qquad |L|=\frac{K}{22.4\sqrt6\sqrt{1.2}}=\frac{K}{60}\ \Rightarrow\ \mathrm{GM}=\frac{60}{20}=3
    2. Gain crossover for K = 20. Square |L| = 1 and solve by trial: the left side is 372 at ω = 12 and 425 at ω = 12.5.

      ω2(1+0.01ω2)(1+0.0004ω2)=400 ⇒ ωgc≈12.3 rad/s\omega^2(1+0.01\omega^2)(1+0.0004\omega^2)=400\ \Rightarrow\ \omega_{\mathrm{gc}}\approx12.3\ \mathrm{rad/s}
    3. Phase margin for K = 20.

      PM=90∘−tan⁡−1(1.227)−tan⁡−1(0.245)=90∘−50.8∘−13.8∘=25.4∘\mathrm{PM}=90^\circ-\tan^{-1}(1.227)-\tan^{-1}(0.245)=90^\circ-50.8^\circ-13.8^\circ=25.4^\circ
    4. For a 45° margin the phase at crossover must be −135°, so the two arctangents must add to 45°. Use the tangent addition formula.

      0.1ω+0.02ω1−0.002ω2=1 ⇒ 0.002ω2+0.12ω−1=0 ⇒ ω=7.42 rad/s\frac{0.1\omega+0.02\omega}{1-0.002\omega^2}=1\ \Rightarrow\ 0.002\omega^2+0.12\omega-1=0\ \Rightarrow\ \omega=7.42\ \mathrm{rad/s}
    5. Choose K so that |L| = 1 at that frequency.

      K=ω1+0.01ω21+0.0004ω2=7.42×1.245×1.011=9.33K=\omega\sqrt{1+0.01\omega^2}\sqrt{1+0.0004\omega^2}=7.42\times1.245\times1.011=9.33
  5. A loop contains a time delay T_d. (a) With T_d = 0 find the gain crossover frequency and the phase margin. (b) Find the largest delay for which the closed loop is still stable. (c) For T_d = 0.1 s find the phase margin and the gain margin.

    L(s)=4 e−sTds (1+0.25s)L(s)=\frac{4\,e^{-sT_d}}{s\,(1+0.25s)}
    Show answer

    Answer(a) ωgc ≈ 3.14 rad/s and PM ≈ 51.8°; without delay the phase never reaches −180°, so the gain margin is infinite. (b) T_d ≈ 0.288 s. (c) PM ≈ 33.8°; the phase now crosses −180° at about 5.93 rad/s, where |L| = 0.377, so GM ≈ 2.65 (8.5 dB).

    PM0≈51.8∘,Td,max⁡≈0.288 s,Td=0.1: PM≈33.8∘, GM≈2.65\mathrm{PM}_0\approx51.8^\circ,\qquad T_{d,\max}\approx0.288\ \mathrm{s},\qquad T_d=0.1:\ \mathrm{PM}\approx33.8^\circ,\ \mathrm{GM}\approx2.65
    Show full solution

    Solution

    1. Gain crossover. The delay does not change the magnitude, so this step is the same for every T_d. With x = ω² the condition |L| = 1 is a quadratic.

      ω2(1+ω216)=16 ⇒ x2+16x−256=0 ⇒ x=9.89,ωgc=3.14 rad/s\omega^2\left(1+\frac{\omega^2}{16}\right)=16\ \Rightarrow\ x^2+16x-256=0\ \Rightarrow\ x=9.89,\quad \omega_{\mathrm{gc}}=3.14\ \mathrm{rad/s}
    2. Phase margin without delay.

      PM0=180∘−90∘−tan⁡−1(0.25×3.14)=90∘−38.2∘=51.8∘\mathrm{PM}_0=180^\circ-90^\circ-\tan^{-1}(0.25\times3.14)=90^\circ-38.2^\circ=51.8^\circ
    3. Largest delay. The delay subtracts ωgc × T_d radians at the crossover. The margin reaches zero when this equals the margin in radians.

      Td,max⁡=51.8∘×π/180∘3.14=0.9053.14=0.288 sT_{d,\max}=\frac{51.8^\circ\times\pi/180^\circ}{3.14}=\frac{0.905}{3.14}=0.288\ \mathrm{s}
    4. Phase margin with T_d = 0.1 s. The extra lag at the crossover is 0.314 rad, which is 18.0°.

      PM=51.8∘−57.3∘×3.14×0.1=51.8∘−18.0∘=33.8∘\mathrm{PM}=51.8^\circ-57.3^\circ\times3.14\times0.1=51.8^\circ-18.0^\circ=33.8^\circ
    5. Gain margin with T_d = 0.1 s. Solve for the phase crossover by trial, then evaluate the magnitude there.

      tan⁡−1(0.25ω)+5.73∘ω=90∘ ⇒ ωpc≈5.93;∣L∣=45.93×1.79=0.377 ⇒ GM=2.65\tan^{-1}(0.25\omega)+5.73^\circ\omega=90^\circ\ \Rightarrow\ \omega_{\mathrm{pc}}\approx5.93;\qquad |L|=\frac{4}{5.93\times1.79}=0.377\ \Rightarrow\ \mathrm{GM}=2.65

Mark each exercise after you check your answer. Marks stay in this browser.

KEY TERMS

Terms this lecture teaches

All terms in the glossary →
QUESTIONS STUDENTS ASK

Frequently asked questions

Why must I convert to Bode form before reading the gain?

A factor such as (s + 2) hides a constant, because it equals 2(1 + s/2). In Bode form every factor equals 1 at low frequency, so the leftover constant is the true low-frequency gain. For 1/(s(s + 1)(s + 2)) that gain is 1/2, and missing it shifts the whole magnitude plot by about 6 dB.

How far is the straight-line asymptote from the true Bode magnitude?

For a first-order pole the true curve is about 3.01 dB below the asymptote at the corner and about 0.97 dB below it one octave to either side. For a quadratic factor the error depends on ζ: at the corner the true magnitude is 1/(2ζ) times the asymptote, so mark the resonant peak of a lightly damped factor first.

Why does crossing 0 dB at −20 dB/dec give a comfortable phase margin?

In a minimum-phase loop the phase is set mainly by the nearby magnitude slope: a slope of −20n dB/dec held for about a decade gives a phase near −90°n. Crossing at −20 dB/dec keeps the phase not far below −90°, while crossing at −40 dB/dec puts it near −180°. Time delays and right-half-plane zeros break this link.

How does a time delay show up on a Bode plot?

A pure delay e^(−sTd) leaves the magnitude plot unchanged and subtracts ωTd radians of phase, about 57.3° × ωTd, a lag that grows without limit with frequency. It lowers the phase margin by that amount at the gain crossover, so the largest tolerable delay is the phase margin in radians divided by the crossover frequency.

Why the next lesson follows

For the simple loops in this chapter, margins read from a Bode plot settle the stability question. The next chapter explains where the pair −180° and 0 dB comes from, namely the point −1 on the Nyquist plot, and gives the test that still works when there are several crossovers or unstable open-loop poles.

YOUR PROGRESS

Finished the reading and the exercises? Mark the lecture as done to tick it on the syllabus. Progress stays in this browser.

EIGENROOM · https://eigenroom.com/theory/bode-construction · Updated 2026-11-15