CHAPTER 30 / FREQUENCY DOMAIN

Designing lead and lag compensators on the Bode plot

Set the gain from the steady-state error, measure how much phase margin is missing, size a lead network from its maximum phase and centre it on the new crossover, or use a lag network to lower the crossover instead. Then check both margins again.

THE SHORT ANSWER

Lead–lag design on the Bode plot first fixes the compensator gain K from the error constant, then reshapes the loop near crossover to meet the phase margin. A lead network K(1 + Ts)/(1 + αTs) supplies at most φm, where sin φm = (1 − α)/(1 + α), and is centred on the new crossover where |KG| = −10 log₁₀(1/α) dB. A lag network instead lowers the high-frequency magnitude so that the loop crosses 0 dB where the plant still has enough phase.

C(s)=K1+Ts1+αTs,sin⁡ϕm=1−α1+α,ωm=1TαC(s)=K\frac{1+Ts}{1+\alpha Ts},\qquad \sin\phi_m=\frac{1-\alpha}{1+\alpha},\qquad \omega_m=\frac{1}{T\sqrt\alpha}
WHAT YOU WILL LEARN

Design a lead compensator on the Bode plot from an error-constant and phase-margin specification, design a lag compensator for the same kind of specification, and verify gain margin, phase margin and the step response afterwards.

The previous lecture designed lead and lag networks by placing closed-loop poles. Many specifications are written instead as an error constant and a phase margin, and for those the Bode plot is the natural workspace: the same networks are sized from their maximum phase and their high-frequency attenuation.

Review the previous lesson: Designing lead and lag compensators on the root locus
START WITH AN EXAMPLE

First, picture it

On a Bode plot the two kinds of specification sit at opposite ends of the frequency axis. The error constant sets the height of the magnitude curve at low frequency, and the phase margin is read where the curve crosses 0 dB. Raising the gain to meet an error requirement lifts the whole magnitude curve, so the crossover moves to the right, where the phase is lower, and the margin shrinks. A lead network repairs this at the crossover. It adds a bump of positive phase, largest at one frequency, so put that frequency at the new crossover. A lag network repairs it the other way. It keeps the low-frequency gain and pulls the high-frequency magnitude down, so the curve crosses 0 dB earlier, at a frequency where the plant still has enough phase. Lead buys margin and speed but raises the high-frequency gain, which amplifies noise; lag buys margin cheaply but makes the loop slower.

THE IDEA

Build the idea step by step

Write each compensator so that its gain at s = 0 is K: C(s) = K(1 + Ts)/(1 + αTs) with α < 1 for lead, and C(s) = K(1 + Ts)/(1 + βTs) with β > 1 for lag. Then K alone sets the error constant. Lead design: (1) choose K from the steady-state error requirement; (2) read the crossover frequency and the phase margin of KG(jω); (3) take the maximum phase φm = required PM − present PM + an allowance of 5° to 12°, because the crossover will move to the right, where the plant has less phase; (4) compute α = (1 − sin φm)/(1 + sin φm); (5) the network adds 10 log10(1/α) dB at its maximum-phase frequency ωm, so the new crossover is where |KG| equals −10 log10(1/α) dB, and ωm goes there; (6) the corners are 1/T = ωm√α and 1/(αT) = ωm/√α; (7) check both margins and repeat with a larger allowance if the phase margin falls short. Lag design: find the frequency where the phase of KG equals −180° + required PM + an allowance of 5° to 12°, and make it the new crossover by attenuating |KG| there to 0 dB with β = |KG|. Put the zero 1/T about a decade below that frequency and the pole at 1/(βT), then check the margins.

Read the equations

Kp=lim⁡s→0C(s)G(s),Kv=lim⁡s→0sC(s)G(s),essramp=1KvK_p=\lim_{s\to0}C(s)G(s),\qquad K_v=\lim_{s\to0}sC(s)G(s),\qquad e_{\mathrm{ss}}^{\mathrm{ramp}}=\frac{1}{K_v}

Error constants of a stable unity-feedback loop. With the compensators written below, C(0) = K, so the steady-state requirement fixes K before anything else.

Clead(s)=K1+Ts1+αTs,0<α<1C_{\mathrm{lead}}(s)=K\frac{1+Ts}{1+\alpha Ts},\qquad 0<\alpha<1

Lead network with gain K at s = 0. Its corners are at 1/T and 1/(αT), and its high-frequency gain is K/α.

sin⁡ϕm=1−α1+α  ⟺  α=1−sin⁡ϕm1+sin⁡ϕm,ωm=1Tα\sin\phi_m=\frac{1-\alpha}{1+\alpha}\iff\alpha=\frac{1-\sin\phi_m}{1+\sin\phi_m},\qquad \omega_m=\frac{1}{T\sqrt\alpha}

Maximum phase φm and the frequency ωm where it occurs, the geometric mean of the two corners.

ϕm=PMreq−PM0+δ,δ≈5∘ to 12∘\phi_m=\mathrm{PM}_{\mathrm{req}}-\mathrm{PM}_0+\delta,\qquad \delta\approx5^\circ\text{ to }12^\circ

Phase the lead must supply. PM0 is the margin of KG alone; the allowance δ covers the phase the plant loses because the crossover moves to the right. Use a larger δ when the plant phase falls steeply above the present crossover.

20log⁡10∣KG(jωc′)∣=−10log⁡101α,ωm=ωc′,1T=ωc′α,1αT=ωc′α20\log_{10}|KG(j\omega_c')|=-10\log_{10}\frac1\alpha,\qquad \omega_m=\omega_c',\qquad \frac1T=\omega_c'\sqrt\alpha,\quad \frac1{\alpha T}=\frac{\omega_c'}{\sqrt\alpha}

New crossover and corner frequencies. The lead adds 10 log10(1/α) dB at ωm, so ωm goes where KG is that much below 0 dB.

Clag(s)=K1+Ts1+βTs,β>1C_{\mathrm{lag}}(s)=K\frac{1+Ts}{1+\beta Ts},\qquad \beta>1

Lag network with gain K at s = 0. Far above its corners it attenuates by 20 log10 β with almost no phase change.

∠KG(jωc′)=−180∘+PMreq+δ,β=∣KG(jωc′)∣,1T=ωc′10,1βT=ωc′10β\angle KG(j\omega_c')=-180^\circ+\mathrm{PM}_{\mathrm{req}}+\delta,\qquad \beta=|KG(j\omega_c')|,\qquad \frac1T=\frac{\omega_c'}{10},\quad \frac1{\beta T}=\frac{\omega_c'}{10\beta}

Lag design. The new crossover comes from the phase curve, β from the magnitude there, and the zero sits a decade below so that the lag costs only about 5.7° at the new crossover.

PM=180∘+∠C(jωc)G(jωc),GM=1∣C(jωpc)G(jωpc)∣\mathrm{PM}=180^\circ+\angle C(j\omega_c)G(j\omega_c),\qquad \mathrm{GM}=\frac{1}{|C(j\omega_{pc})G(j\omega_{pc})|}

Verification. ωc is where |CG| = 1 and ωpc is where ∠CG = −180°. Recompute both on the compensated loop; the design is done only when both meet the specification.

What the symbols mean

G(s)G(s)
plant transfer function in unity feedback
KK
compensator gain at s = 0, set by the error constant
T, α, βT,\ \alpha,\ \beta
time constant of the compensator zero (s), lead ratio α < 1 and lag ratio β > 1
ϕm, ωm\phi_m,\ \omega_m
maximum phase of the lead network (degrees) and the frequency where it occurs (rad/s)
ωc, ωc′\omega_c,\ \omega_c'
gain-crossover frequency before compensation and the new one after, rad/s
ωpc\omega_{pc}
phase-crossover frequency, where the loop phase is −180°, rad/s
PM, PM0, PMreq\mathrm{PM},\ \mathrm{PM}_0,\ \mathrm{PM}_{\mathrm{req}}
phase margin of the compensated loop, of KG alone, and the required value, degrees
GM\mathrm{GM}
gain margin, as a ratio or in dB
δ\delta
phase allowance for the shift of the crossover, degrees
Kp, KvK_p,\ K_v
position and velocity error constants; the ramp error is 1/Kv

The margin rules here are for a loop that is stable in open loop apart from integrators, minimum phase, and crosses 0 dB once; for such a loop positive margins mean a stable closed loop. Margins are not step-response specifications. The compensator's zero is also a closed-loop zero and can add overshoot, so simulate the step as well.

BODE PLOT / gain alone, lead and lag
0.010.111010010000 dB−180°|L| dB∠Lω (rad/s)

Grey: KG with K = 20, crossover 6.14 rad/s and phase margin 11.0°. Blue: with the lead, crossover 9.76 rad/s and margin 46.5°. Amber, dashed: the lag alternative from the “More detail” section, crossover 1.53 rad/s and margin 45.8°. The thick vertical bars are the phase margins.

WORKED EXAMPLE

Lead design for G(s) = 1/(s(0.5s+1)(0.02s+1)) with Kv = 20 s⁻¹

  1. Specifications. Unity feedback, Kv = 20 s⁻¹ (a ramp error of 5%), PM ≥ 45° and GM ≥ 10 dB. With C(s) = K(1 + Ts)/(1 + αTs), Kv equals K.

    Kv=lim⁡s→0s C(s)G(s)=K=20K_v=\lim_{s\to0}s\,C(s)G(s)=K=20
  2. Uncompensated margins. |KG| crosses 0 dB at 6.14 rad/s, where the phase is −169.0°, so PM = 11.0°. The phase reaches −180° at exactly 10 rad/s, because 0.5ω × 0.02ω = 1 there, and |KG(j10)| = 0.385, so GM = 8.3 dB. Both margins fail.

    ωc=6.14, PM0=11.0∘;ωpc=10, GM=10.385=2.6=8.3 dB\omega_c=6.14,\ \mathrm{PM}_0=11.0^\circ;\qquad \omega_{pc}=10,\ \mathrm{GM}=\frac{1}{0.385}=2.6=8.3\ \mathrm{dB}
  3. Allowance. Between 6.14 and 10 rad/s the plant phase falls by 11°, and the lead will push the crossover into that range. Take δ = 12°.

    ϕm=45∘−11.0∘+12∘=46∘\phi_m=45^\circ-11.0^\circ+12^\circ=46^\circ
  4. Lead ratio.

    α=1−sin⁡46∘1+sin⁡46∘=0.163,10log⁡101α=7.87 dB\alpha=\frac{1-\sin46^\circ}{1+\sin46^\circ}=0.163,\qquad 10\log_{10}\frac1\alpha=7.87\ \mathrm{dB}
  5. New crossover. On the uncompensated curve, |KG| = −7.87 dB at 9.76 rad/s. Put ωm there.

    20log⁡10∣KG(j9.76)∣=−7.87 dB ⇒ ωm=ωc′=9.76 rad/s20\log_{10}|KG(j9.76)|=-7.87\ \mathrm{dB}\ \Rightarrow\ \omega_m=\omega_c'=9.76\ \mathrm{rad/s}
  6. Corner frequencies and the compensator.

    1T=9.760.163=3.94,1αT=9.760.163=24.1;C(s)=20 1+0.254s1+0.0414s\frac1T=9.76\sqrt{0.163}=3.94,\quad \frac1{\alpha T}=\frac{9.76}{\sqrt{0.163}}=24.1;\qquad C(s)=20\,\frac{1+0.254s}{1+0.0414s}
  7. Check the phase margin. At 9.76 rad/s the lead adds +7.87 dB and +46.0°, so |CG| = 1 and the phase is the plant's −179.46° plus 46.0°.

    ∠CG=−90∘−78.42∘−11.04∘+46.00∘=−133.46∘,PM=46.5∘≥45∘\angle CG=-90^\circ-78.42^\circ-11.04^\circ+46.00^\circ=-133.46^\circ,\qquad \mathrm{PM}=46.5^\circ\ge45^\circ
  8. Check the gain margin with a frequency sweep of the compensated loop: the phase reaches −180° at 32.6 rad/s, where |CG| = 0.156, so GM = 16.1 dB.

    ωpc=32.6 rad/s,GM=10.156=6.42=16.1 dB≥10 dB\omega_{pc}=32.6\ \mathrm{rad/s},\qquad \mathrm{GM}=\frac{1}{0.156}=6.42=16.1\ \mathrm{dB}\ge10\ \mathrm{dB}
  9. Closed loop and step response. The closed-loop poles are −4.99, −7.34 ± j10.83 and −56.5, with a zero at −3.94. A step simulation gives 25.3% overshoot and a 2% settling time of 0.53 s. With the same Kv and no lead, the loop overshoots 73.5% and needs 6.6 s.

    Mp=25.3%, ts=0.53 s(without lead: 73.5%, 6.6 s)M_p=25.3\%,\ t_s=0.53\ \mathrm{s}\qquad(\text{without lead: }73.5\%,\ 6.6\ \mathrm{s})
REFERENCE

Reference table

Sizing a lead network K(1 + Ts)/(1 + αTs) from its maximum phase φm
φm (°)α\alphaCorner ratio 1/αGain added at ωm (dB)High-frequency gain rise (dB)
200.49032.043.106.19
300.33333.004.779.54
450.17165.837.6615.31
500.13257.558.7817.56
550.099410.0610.0320.05
600.071813.9311.4422.88
650.049120.3513.0826.17
700.031132.1615.0730.15

α = (1 − sin φm)/(1 + sin φm). Put ωm = 1/(T√α) where |KG| = −10 log₁₀(1/α) dB. Sensor noise reaches the actuator amplified by the high-frequency gain rise.

More detail and model assumptions
A CLOSER LOOK

Why the gain comes first

The steady-state error of a stable unity-feedback loop depends only on the loop gain at very low frequency: Kp = C(0)G(0) for a step into a type-0 loop and Kv = lim sC(s)G(s) for a ramp into a type-1 loop. Writing the lead and lag networks with (1 + Ts) factors makes them equal to 1 at s = 0, so the K in front of them is the only thing that affects the error constant. That is why K is fixed first and never changed afterwards; the network is then shaped around the crossover without touching the low-frequency asymptote. The root-locus form K_c(s + z)/(s + p) describes the same network, but its value at s = 0 is K_c z/p, so convert before comparing the two lectures.

A CLOSER LOOK

Where the maximum phase and the −10 log10(1/α) come from

The phase of (1 + jωT)/(1 + jωαT) is tan⁻¹(ωT) − tan⁻¹(ωαT). Setting its derivative to zero gives ωm = 1/(T√α), the geometric mean of the two corner frequencies 1/T and 1/(αT), which sits in the middle of them on the logarithmic axis. Substituting ωm back gives sin φm = (1 − α)/(1 + α), and the magnitude there is 1/√α, which is 10 log10(1/α) dB, exactly half of the total high-frequency increase 20 log10(1/α). So if ωm is to be the new crossover, the uncompensated KG must be −10 log10(1/α) dB there: the network's +10 log10(1/α) dB lifts it to exactly 0 dB.

The formula also shows the practical ceiling. φm = 60° needs 1/α ≈ 14, a 23 dB rise in high-frequency gain; beyond that each extra degree costs a great deal of noise gain. If more phase is needed, use two lead stages, or a lag-lead combination that first lowers the crossover.

A CLOSER LOOK

Lag design on the phase curve

A lag network has (nearly) no phase effect far above its corners and attenuates the magnitude there by 20 log10 β. So pick the new crossover from the phase curve of KG, where the phase is −180° + required PM + allowance, and choose β to bring |KG| down to 0 dB at that frequency. The allowance pays for the lag's own phase at the new crossover. With the zero one decade below, that lag is about tan⁻¹(0.1) = 5.7°, so an allowance of 5° is not quite enough.

Applied to the worked example's plant with K = 20: with a 6° allowance the phase of KG is −129° at 1.52 rad/s, where |KG| = 10.46 (20.4 dB). The lag C(s) = 20(1 + 6.57s)/(1 + 68.8s) gives PM = 45.8° and GM = 28.0 dB. With a 5° allowance the same procedure ends at PM = 44.8°, just short of 45°. The lag design meets the margins with a crossover of 1.53 rad/s, more than six times lower than the lead design's 9.76 rad/s. Its simulated step overshoots 26.4% and needs 8.7 s to settle within 2%, against 0.53 s for the lead design.

A CLOSER LOOK

Choosing lead or lag, and when neither is enough

Choose lead when the closed loop must get faster and the plant's phase falls slowly just above the present crossover, so that a moderate α reaches the margin. Choose lag when the present speed is acceptable, the plant has enough phase at a lower frequency, and noise at high frequency matters. When the plant's phase falls steeply above the crossover, each degree of lead moves the crossover into territory with even less phase, and the required α becomes impractically small (exercise 5). Then combine the two: a lag stage lowers the crossover part of the way and a lead stage supplies the rest. A lead whose pole moves to infinity becomes PD control, and a lag whose pole moves to the origin becomes PI control, which is where the next part of the course begins.

TRY IT

Connect the example to the lab

The motor lab's lead lesson starts with C(s) = 2(s + 4)/(s + 8). Open the frequency workbench: it reports a phase margin of 48.7° at a gain crossover of 17.10 rad/s. Now move only the zero up to 7.95 rad/s, right next to the pole. The network becomes almost a plain gain of 2, and the workbench reports 35.6° at 17.46 rad/s. So the lead's 4–8 rad/s band adds about 13° of margin at nearly the same crossover. The workbench analyses the sampled loop with its 20 ms hold, so its phase falls faster at high frequency than a continuous-time sketch would suggest.

Read the lead network's phase margin on the motor
YOUR EXPERIMENT

Make a prediction before moving a control.

  1. PREDICTIn the lead lesson the gain is 2 and the pole is at 8 rad/s. Predict whether the phase margin rises or falls as the zero moves from 4 rad/s toward the pole, and whether the gain crossover moves much.
  2. CHANGE ONE THINGKeep the gain at 2 and the pole at 8 rad/s, set the zero to 4, 6 and then 7.95 rad/s, and pin each run.
  3. OBSERVERead the phase margin, gain margin and gain crossover in the frequency workbench, and compare the overshoot and final error of the pinned runs.
  4. EXPLAINUse the maximum-phase formula to explain the trend: a larger pole-to-zero ratio gives more phase. Then explain why the final error falls as the zero moves up, using the network's gain 2z/p at s = 0.
BE CAREFUL

A phase margin is not an overshoot specification. In the worked example PM = 46.5° suggests ζ ≈ 0.47 and about 19% overshoot by the usual rule of thumb, but the simulated step overshoots 25.3%, because the lead zero at −3.94 becomes a closed-loop zero near a slow real pole at −4.99. Also remember the cost: the lead raises the high-frequency loop gain by 20 log10(1/α) dB, 15.7 dB in the example, and sensor noise reaches the actuator amplified by the same factor.

Check your understanding+

After a lead design, the computed phase margin is 2° short of the requirement. Which step do you repeat, and which way do you change it?

Increase the allowance, and with it φm, recompute α, find the new crossover where |KG| = −10 log10(1/α) dB, and check again. A larger φm pushes the crossover further right, where the plant has less phase, so each extra degree of φm buys less than a degree of margin: in exercise 3, going from 30° to 32° raises the margin from 49.3° to 50.9°.
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. A unity-feedback loop has the plant below. The steady-state error to a unit ramp must not exceed 0.025. Find the smallest gain K, the gain-crossover frequency and phase margin of KG, and the gain margin. What overshoot does the closed loop have?

    G(s)=1s(0.25s+1)G(s)=\frac{1}{s(0.25s+1)}
    Show answer

    AnswerK = 40. The crossover is at 12.3 rad/s and PM = 18.0°. The gain margin is infinite, because the phase approaches −180° but never reaches it. Even so, the closed loop has ζ = 0.158 and overshoots 60.5%.

    K=40,ωc=12.3 rad/s,PM=18.0∘,GM=∞K=40,\quad \omega_c=12.3\ \mathrm{rad/s},\quad \mathrm{PM}=18.0^\circ,\quad \mathrm{GM}=\infty
    Show full solution

    Solution

    1. The loop is type 1, so the ramp error is 1/Kv and Kv = K.

      Kv=lim⁡s→0sKs(0.25s+1)=K≥10.025=40K_v=\lim_{s\to0}s\frac{K}{s(0.25s+1)}=K\ge\frac1{0.025}=40
    2. Crossover. Set |KG| = 1 and write x = (0.25ω)², so that ω² = 16x.

      40ω1+(0.25ω)2=1⇒x(1+x)=100, x=9.51⇒ωc=9.510.25=12.3 rad/s\frac{40}{\omega\sqrt{1+(0.25\omega)^2}}=1\Rightarrow x(1+x)=100,\ x=9.51\Rightarrow\omega_c=\frac{\sqrt{9.51}}{0.25}=12.3\ \mathrm{rad/s}
    3. Phase margin.

      PM=180∘−90∘−tan⁡−1(0.25×12.3)=90∘−72.0∘=18.0∘\mathrm{PM}=180^\circ-90^\circ-\tan^{-1}(0.25\times12.3)=90^\circ-72.0^\circ=18.0^\circ
    4. Gain margin and step. The phase −90° − tan⁻¹(0.25ω) never reaches −180°, so GM is infinite, yet the closed-loop poles of s² + 4s + 160 are lightly damped.

      0.25s2+s+40=0  ⟺  s2+4s+160=0,ζ=42160=0.158,Mp=60.5%0.25s^2+s+40=0\iff s^2+4s+160=0,\quad \zeta=\frac{4}{2\sqrt{160}}=0.158,\quad M_p=60.5\%
  2. A lead network must supply its maximum phase of 40° at 8 rad/s. Find α, the gain it adds at 8 rad/s, its two corner frequencies and the increase in high-frequency gain.

    Show answer

    Answerα = 0.217. At 8 rad/s the network adds 6.63 dB. The corners are at 3.73 and 17.2 rad/s. The high-frequency gain rises by 1/α = 4.60, or 13.3 dB.

    1+0.268s1+0.0583s\frac{1+0.268s}{1+0.0583s}
    Show full solution

    Solution

    1. Lead ratio from the maximum phase.

      α=1−sin⁡40∘1+sin⁡40∘=1−0.6431+0.643=0.217\alpha=\frac{1-\sin40^\circ}{1+\sin40^\circ}=\frac{1-0.643}{1+0.643}=0.217
    2. Gain added at ωm, half of the total high-frequency increase in dB.

      10log⁡101α=10log⁡104.60=6.63 dB10\log_{10}\frac1\alpha=10\log_{10}4.60=6.63\ \mathrm{dB}
    3. Corners on either side of ωm, at equal distances on the logarithmic axis.

      1T=ωmα=8×0.466=3.73 rad/s,1αT=ωmα=17.2 rad/s\frac1T=\omega_m\sqrt\alpha=8\times0.466=3.73\ \mathrm{rad/s},\qquad\frac1{\alpha T}=\frac{\omega_m}{\sqrt\alpha}=17.2\ \mathrm{rad/s}
    4. High frequencies. The ratio (1 + jωT)/(1 + jωαT) tends to 1/α, so sensor noise above about 17 rad/s reaches the actuator 4.6 times stronger than it would through the gain K alone.

      20log⁡101α=13.3 dB20\log_{10}\frac1\alpha=13.3\ \mathrm{dB}
  3. Design a lead compensator for the plant below so that Kv = 50 s⁻¹ and PM ≥ 50°. Start with a 5° allowance, check the result, and correct it if necessary.

    G(s)=1s(0.1s+1)G(s)=\frac{1}{s(0.1s+1)}
    Show answer

    AnswerK = 50 gives PM = 25.2° at 21.3 rad/s. A first try with φm = 30° (α = 1/3) gives only 49.3°. A second try with φm = 32° (α = 0.307) puts the crossover at 29.2 rad/s with corners at 16.2 and 52.7 rad/s and gives PM = 50.9°.

    C(s)=50 1+0.0618s1+0.0190s,PM=50.9∘C(s)=50\,\frac{1+0.0618s}{1+0.0190s},\qquad \mathrm{PM}=50.9^\circ
    Show full solution

    Solution

    1. Gain and uncompensated margin. Write x = (0.1ω)², so that ω² = 100x.

      K=50:x(1+x)=25⇒x=4.52, ωc=21.3,PM0=90∘−tan⁡−12.13=25.2∘K=50:\quad x(1+x)=25\Rightarrow x=4.52,\ \omega_c=21.3,\quad \mathrm{PM}_0=90^\circ-\tan^{-1}2.13=25.2^\circ
    2. First try: φm = 50° − 25.2° + 5° ≈ 30°, so α = 1/3 and the new crossover is where |KG| = −4.77 dB, that is |KG| = 1/√3.

      x(1+x)=75⇒ωc′=28.6,PM=90∘−tan⁡−12.86+30∘=49.3∘<50∘x(1+x)=75\Rightarrow\omega_c'=28.6,\qquad \mathrm{PM}=90^\circ-\tan^{-1}2.86+30^\circ=49.3^\circ<50^\circ
    3. The plant lost 5.9° between 21.3 and 28.6 rad/s, more than the 5° allowance. Second try: φm = 32°.

      α=1−sin⁡32∘1+sin⁡32∘=0.307,x(1+x)=250.307⇒ωc′=29.2,PM=90∘−tan⁡−12.92+32∘=50.9∘\alpha=\frac{1-\sin32^\circ}{1+\sin32^\circ}=0.307,\quad x(1+x)=\frac{25}{0.307}\Rightarrow\omega_c'=29.2,\quad \mathrm{PM}=90^\circ-\tan^{-1}2.92+32^\circ=50.9^\circ
    4. Corners and the compensator.

      1T=29.20.307=16.2,1αT=29.20.307=52.7,C(s)=50 1+0.0618s1+0.0190s\frac1T=29.2\sqrt{0.307}=16.2,\quad\frac1{\alpha T}=\frac{29.2}{\sqrt{0.307}}=52.7,\qquad C(s)=50\,\frac{1+0.0618s}{1+0.0190s}
    5. Check. The phase of this loop never reaches −180°, so the gain margin is infinite. A step simulation gives 19.5% overshoot and a 2% settling time of 0.17 s.

  4. Design a lag compensator for the plant below with Kv = 10 s⁻¹, PM ≥ 45° and GM ≥ 10 dB. Use a 6° allowance and put the zero one decade below the new crossover.

    G(s)=1s(0.5s+1)(0.1s+1)G(s)=\frac{1}{s(0.5s+1)(0.1s+1)}
    Show answer

    AnswerK = 10; without compensation PM = 3.9° and GM = 1.6 dB. The new crossover is 1.245 rad/s, β = 6.77 (16.6 dB), the zero is at 0.125 rad/s and the pole at 0.0184 rad/s. The result has PM = 46.0° and GM = 17.6 dB.

    C(s)=10 1+8.03s1+54.3sC(s)=10\,\frac{1+8.03s}{1+54.3s}
    Show full solution

    Solution

    1. Gain and uncompensated margins. K = Kv = 10. The phase reaches −180° where 0.5ω × 0.1ω = 1, at √20 = 4.47 rad/s, and |KG| = 0.833 there; the magnitude crosses 0 dB at 4.08 rad/s.

      GM0=10.833=1.2=1.6 dB,PM0=3.9∘\mathrm{GM}_0=\frac{1}{0.833}=1.2=1.6\ \mathrm{dB},\qquad \mathrm{PM}_0=3.9^\circ
    2. Target phase −180° + 45° + 6° = −129°, so the two plant lags must total 39°. The tangent addition formula turns this into a quadratic in ω.

      tan⁡−10.5ω+tan⁡−10.1ω=39∘⇒0.6ω1−0.05ω2=tan⁡39∘⇒ωc′=1.245 rad/s\tan^{-1}0.5\omega+\tan^{-1}0.1\omega=39^\circ\Rightarrow\frac{0.6\omega}{1-0.05\omega^2}=\tan39^\circ\Rightarrow\omega_c'=1.245\ \mathrm{rad/s}
    3. Attenuation needed at the new crossover.

      β=∣KG(j1.245)∣=101.2451+0.388 1+0.0155=6.77 (16.6 dB)\beta=|KG(j1.245)|=\frac{10}{1.245\sqrt{1+0.388}\,\sqrt{1+0.0155}}=6.77\ (16.6\ \mathrm{dB})
    4. Zero and pole.

      1T=1.24510=0.125,1βT=0.1256.77=0.0184,C(s)=10 1+8.03s1+54.3s\frac1T=\frac{1.245}{10}=0.125,\quad \frac1{\beta T}=\frac{0.125}{6.77}=0.0184,\qquad C(s)=10\,\frac{1+8.03s}{1+54.3s}
    5. Verify with a frequency sweep: PM = 46.0° at 1.250 rad/s, and the phase reaches −180° at 4.33 rad/s with GM = 17.6 dB. A step simulation overshoots 25.9% and needs 10.8 s to settle within 2%: the price of a crossover more than three times lower than before.

  5. Return to the worked example's plant but demand Kv = 50 s⁻¹ with PM ≥ 45°. Show how far a single lead network must be pushed, and compare it with a lag design.

    G(s)=1s(0.5s+1)(0.02s+1)G(s)=\frac{1}{s(0.5s+1)(0.02s+1)}
    Show answer

    AnswerWith K = 50 the loop is almost at the stability limit: PM = 0.4° and GM = 0.34 dB. A lead with φm = 55° reaches only 42.6°; φm = 60° just reaches 45.7° but needs 1/α = 13.9 and raises the high-frequency gain by 22.9 dB. A lag with a 6° allowance gives PM = 45.4° with β = 26.1, but lowers the crossover to 1.52 rad/s. Neither is attractive; a lag-lead compensator or a smaller Kv is the usual way out.

    Show full solution

    Solution

    1. Uncompensated margins with K = 50. The phase crossover is still 10 rad/s, but now |KG(j10)| = 0.962.

      ωc=9.81 rad/s,PM0=0.4∘,GM=10.962=1.04=0.34 dB\omega_c=9.81\ \mathrm{rad/s},\quad \mathrm{PM}_0=0.4^\circ,\qquad \mathrm{GM}=\frac{1}{0.962}=1.04=0.34\ \mathrm{dB}
    2. Phase budget. Above 10 rad/s the plant phase keeps falling: −180.0° at 10 rad/s and −196.1° at 20 rad/s. Every degree of lead moves the crossover into this region.

      ∠G(j10)=−180.0∘,∠G(j20)=−196.1∘\angle G(j10)=-180.0^\circ,\qquad \angle G(j20)=-196.1^\circ
    3. Lead tries, each checked with a frequency sweep: φm = 55° puts the crossover at 17.3 rad/s with PM = 42.6°; φm = 60° puts it at 18.6 rad/s with PM = 45.7°.

      ϕm=60∘:1α=13.9,20log⁡101α=22.9 dB\phi_m=60^\circ:\quad \frac1\alpha=13.9,\qquad 20\log_{10}\frac1\alpha=22.9\ \mathrm{dB}
    4. Lag alternative. With a 6° allowance the phase of KG is −129° at 1.52 rad/s, where |KG| = 26.1 (28.3 dB). With the zero at 0.152 rad/s the result is PM = 45.4° and GM = 28.0 dB, but the crossover is more than six times lower than the uncompensated 9.81 rad/s.

    5. Conclusion. When the plant phase falls this steeply above the crossover, combine the two: a lag stage that lowers the crossover part of the way and a lead stage that supplies the rest. Or ask whether Kv = 50 is really needed.

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 is the gain K chosen before the lead or lag network?

Written as K(1 + Ts)/(1 + αTs) or K(1 + Ts)/(1 + βTs), the compensator equals K at s = 0, so K alone sets the error constant. Fixing it first lets the network reshape only the crossover region without disturbing the steady-state requirement.

Why add 5° to 12° of extra phase in a lead design?

The lead also raises the magnitude, so the crossover moves to a higher frequency where the plant has less phase. The allowance pays for that loss. Take more when the plant phase falls steeply above the old crossover, and repeat with a larger allowance if the margin comes out short.

How much phase can one lead network provide?

In principle anything below 90°, but the cost grows quickly: 45° needs a corner ratio 1/α ≈ 5.83 and raises high-frequency gain by 15.31 dB, while 60° needs about 13.93 and 22.88 dB. Beyond roughly 60°, use two lead stages or a lag–lead combination instead.

Does a phase margin guarantee a certain overshoot?

No. The rule of thumb linking phase margin to damping holds for a pure second-order loop. A lead compensator's zero is also a closed-loop zero and can add overshoot, so the step response of a lead design must be simulated rather than read from the margin.

When should I choose lag instead of lead?

Choose lag when the present speed is acceptable, the plant has enough phase at a lower frequency and high-frequency noise matters: it buys margin by lowering the crossover, so the loop gets slower. Choose lead when the loop must get faster and the plant phase falls slowly above the crossover; when it falls steeply, combine the two.

Why the next lesson follows

Lead and lag networks are close relatives of PD and PI control: a lead whose pole moves to infinity becomes PD, and a lag whose pole moves to the origin becomes PI. The next part turns to PID, the controller used in most real loops.

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/lead-lag-frequency · Updated 2026-11-15