FREQUENCY-DOMAIN PRACTICE 7 MIN READ

Why a lead adds less phase margin than its φmax

A lead network peaks at one frequency and lifts the loop gain, so the crossover moves to where the plant has less phase. Here is what that costs, in degrees.

You need about 20° more phase margin, so you add a lead network whose maximum phase is 36.9°, and the margin rises by 10.9°. The formula for φmax is not wrong. The network reaches that phase at one frequency only, and it also raises the loop gain, which moves the crossover to a frequency where the plant has less phase to spare.

01 / FIELD NOTE

Where the missing degrees go

Write the lead with unit gain at DC: C(s) = (1 + s/z)/(1 + s/p) with z < p and α = z/p. Its phase peaks at ωm = √(zp), where it equals φmax = arcsin((1 − α)/(1 + α)), and there its magnitude is 1/√α, that is 10·log10(1/α) dB. Above ωm the gain keeps rising towards 1/α while the phase falls back towards zero.

Because |C| > 1 above the old crossover ωc, the compensated loop crosses 0 dB later, at some ωc′ > ωc. The new margin is the old one plus two terms: PM′ = PM + ∠C(jωc′) + [∠G(jωc′) − ∠G(jωc)]. The first term is at most φmax and equals it only if ωc′ happens to be ωm. The second is the plant phase lost between the two crossovers, and near crossover the plant phase almost always falls with frequency.

02 / FIELD NOTE

One plant, three placements

Take G(s) = 120/(s(s+2)(s+10)), a type 1 loop with Kv = 6 s⁻¹. It crosses 0 dB at 3.10 rad/s with a phase margin of 15.6°, and its phase reaches −180° at 4.47 rad/s, which gives a gain margin of 6.0 dB. The lead C(s) = 4(s+1.5)/(s+6) has α = 0.25 and φmax = 36.9° at ωm = 3.0 rad/s, almost exactly on the old crossover, and unit DC gain, so Kv stays at 6 s⁻¹.

Centred on the old crossover, the lead adds 33.0° at the new crossover of 5.21 rad/s, but the plant phase there is 22.0° lower than at 3.10 rad/s, so the margin rises only from 15.6° to 26.5°: 10.9° of the 36.9° on paper. Halving the lead's gain keeps the crossover near 3.16 rad/s and the margin becomes 51.7°, almost the full φmax, at the price of half the velocity constant. Moving ωm to 4.47 rad/s, where the uncompensated |G| is −6.0 dB, makes ωm the new crossover; the lead then supplies all 36.9°, and the margin is 36.9° because the plant phase there is already −180°.

PlacementCrossover (rad/s)Lead phase therePlant phase changePhase marginGain margin
No lead3.10——15.6°6.0 dB
4(s+1.5)/(s+6), on the old crossover5.2133.0°−22.0°26.5°7.2 dB
Same lead at half gain (Kv = 3 s⁻¹)3.1636.8°−0.7°51.7°13.2 dB
4(s+2.236)/(s+8.944), on 4.47 rad/s4.4736.9°−15.6°36.9°10.5 dB
03 / FIELD NOTE

Try it

In the Bode plot calculator, enter Numerator 120 and Denominator s(s+2)(s+10) with K = 1 and read the phase margin of 15.6° at 3.10 rad/s. Change them to 480(s+1.5) and s(s+2)(s+6)(s+10) for the lead centred on the old crossover (26.5° at 5.21 rad/s), then set K = 0.5 (51.7° at 3.16 rad/s). The links below this note open each case.

On the motor, open the Lead experiment (K = 2, z = 4 rad/s, p = 8 rad/s, so φmax = 19.5° at ωm = 5.66 rad/s) and the frequency workbench. The phase margin is 48.7° at a gain crossover of 17.1 rad/s, three times ωm, where this lead adds only about 11.9°. Raise the pole frequency p to 24 rad/s, then the zero frequency z to 12 rad/s: the same ratio puts ωm at 17.0 rad/s, and the margin becomes 66.9° at 14.0 rad/s. With Target 5 rad/s and Load torque 0, so that the voltage stays below 12 V, the first setting rises 12.2% above the target and the second not at all.

Compared with what, though? On the same sampled motor model, a plain gain of 2 (the lead's high-frequency gain) gives 35.5° at 17.5 rad/s, and a plain gain of 1 (its DC gain) gives 54.8° at 10.7 rad/s. The lesson's lead is 13.2° better than the first and 6.2° worse than the second. The P slider stops at 0.8, so these two numbers come from the workbench's loop model rather than from a run you can repeat. A lead earns its phase only against the gain it is meant to replace.

04 / FIELD NOTE

What these numbers do not show

The calculator works on the continuous, linear loop. The motor workbench uses the 20 ms sampled loop and leaves out load torque, sensor noise and the ±12 V limit. At the lesson's own 120 rad/s target both lead settings saturate and neither overshoots, so the margin is not what shapes that run. Phase margin predicts overshoot only roughly, through the dominant poles; the 12.2% here is a measured fact for one step size, not a rule.

Design procedures add 5° to 12° to the required φmax for exactly the plant-phase term above. In this example the term was 15.6°, beyond the usual allowance, because the poles at 2 and 10 rad/s make the phase fall steeply between 3 and 4.5 rad/s. Read the plant's phase slope near the intended crossover before choosing α.

TAKEAWAY

A lead's φmax is an upper bound at one frequency. The margin you get is φmax, minus what the lead gives up away from ωm, minus the plant phase lost as its extra gain moves the crossover.

CHECK YOUR UNDERSTANDING: A lead with α = 0.1 has φmax = 54.9°. Your loop needs 40° more phase margin. Is this lead enough?

That number alone cannot tell you. The lead raises the gain by 10 dB at ωm, so the crossover moves up. Compute the lead's phase and the plant's phase at the new crossover, or keep the gain down and accept a lower error constant.