Second-Order System Response

Why second-order response matters

Almost every dynamic system an engineer designs boils down, near its dominant behavior, to a second-order model. The reason is physical: a second-order system is one that stores energy in two independent places and trades that energy back and forth. A mass stores kinetic energy while a spring stores potential energy; an inductor stores energy in its magnetic field while a capacitor stores it in its electric field. Whenever two such reservoirs are coupled, the system can oscillate, and the mathematics is identical whether the medium is mechanical, electrical, thermal, or hydraulic.

• Control loops. A well-tuned closed loop is deliberately shaped to look second-order with ζ≈0.5–0.8, so overshoot and settling time are predictable.
• RLC circuits. A series RLC has ωn = 1/√(LC) and ζ = (R/2)·√(C/L); it is the electrical twin of the spring-mass-damper.
• Vehicle suspension. A quarter-car spring and shock absorber target ζ≈0.25–0.35 for ride comfort, trading a little overshoot for a soft response.
• Servo and motion systems. Position loops are tuned so the closed-loop poles form a dominant second-order pair setting bandwidth and overshoot.
• Phase-locked loops. A PLL's loop filter makes the phase error a second-order system, with ζ and ωn setting lock time and jitter.
• Instrumentation. Accelerometers, galvanometers, and pressure sensors are second-order; their damping sets bandwidth and ringing.

Because one model covers so much ground, the closed-form formulas for overshoot, rise time, and settling time are among the most-used equations in all of engineering. Learn them once and you can specify a suspension, a filter, and a robot joint with the same arithmetic.

How it works, step by step

The standard second-order transfer function relating output Y(s) to input X(s) is

G(s) = Y(s)/X(s) = ωn² / (s² + 2ζωn·s + ωn²)

where the symbols are:

• s — the Laplace variable, in rad/s (a complex frequency).
• ωn — undamped natural frequency, in rad/s. It is the frequency at which the system would ring if there were no damping.
• ζ (zeta) — damping ratio, dimensionless. The ratio of actual damping to the critical damping that just prevents oscillation.

The DC gain here is 1 (set numerator to ωn² and s→0), so the output settles to the input value; a physical gain K simply multiplies the whole thing. The character of the response comes from the roots of the denominator — the poles:

s = −ζωn ± jωn·√(1 − ζ²)

Read those poles geometrically. Their distance from the origin is always ωn. Their horizontal position (real part) is σ = −ζωn, which controls decay. Their vertical position (imaginary part) is the damped natural frequency ωd = ωn·√(1−ζ²), the frequency the system actually rings at. The angle β = arccos(ζ) that the pole makes with the negative real axis is set entirely by ζ. So ζ steers the pole angle and ωn its radius.

1. Apply a unit step. The input jumps from 0 to 1 and stays there. The system must move from its old equilibrium to the new one.
2. The poles decide the path. Complex poles (ζ<1) mean the output overshoots and rings at ωd inside a decaying envelope e^(−ζωn·t). Real poles (ζ≥1) mean a smooth, non-oscillating climb.
3. The envelope sets settling. The oscillation amplitude is squeezed by e^(−ζωn·t); once it drops below the tolerance band, the system is "settled." Since σ = ζωn, farther-left poles settle faster.
4. ζ sets overshoot, ωn sets speed. Change ωn and the whole waveform stretches or compresses in time but keeps its shape and overshoot. Change ζ and the shape itself changes.

For the underdamped case, the closed-form unit-step response is y(t) = 1 − e^(−ζωn·t)·[cos(ωd·t) + (ζ/√(1−ζ²))·sin(ωd·t)], which peaks above 1 and rings down to 1.

The three damping regimes

The single parameter ζ divides all second-order behavior into distinct regimes. The table below lists the pole structure and step-response character for each, with representative time-domain metrics for a fixed ωn.

Regime	ζ range	Poles	Step response	Overshoot
Undamped	ζ = 0	±jωn on imaginary axis	Rings forever, never settles	100%
Underdamped	0 < ζ < 1	Complex conjugate pair	Overshoots, then rings down	0–100%
Critically damped	ζ = 1	Repeated real pole at −ωn	Fastest with no overshoot	0%
Overdamped	ζ > 1	Two distinct real poles	Slow, sluggish, no overshoot	0%

The most important design insight hides in this table: critically damped is the fastest response that does not overshoot. Beginners often assume "more damping is safer, so crank ζ up." But once ζ passes 1, a slow real pole appears and dominates, so the overdamped system is slower than the critically damped one while gaining nothing. Meanwhile a slightly underdamped design at ζ≈0.7 rises faster than either and pays only about 4% overshoot for it.

Transient metrics and where the numbers come from

For an underdamped second-order system responding to a step, four metrics fully characterize the transient. Each follows directly from ζ and ωn:

• Percent overshoot: %OS = 100·e^(−ζπ/√(1−ζ²)). Depends on ζ only — not on ωn.
• Peak time: Tp = π/ωd = π/(ωn√(1−ζ²)). Time to the first peak.
• Rise time (0–100%): Tr = (π − β)/ωd, with β = arccos(ζ). A handy 10–90% approximation is Tr ≈ 1.8/ωn near ζ=0.5.
• Settling time (2%): Ts ≈ 4/(ζωn); the 5% band gives Ts ≈ 3/(ζωn).

Notice the clean division of labor. Overshoot is a function of ζ alone. Settling time depends on the product ζωn, which is exactly σ, the real part of the poles. Peak and rise time scale as 1/ωn. This is why the design recipe is simple: pick ζ from the overshoot you can tolerate, then pick ωn from the speed you need. The table below shows the overshoot that different damping ratios produce.

Damping ratio ζ	Percent overshoot	Regime / note
0.1	≈ 73%	Very lightly damped, long ringing
0.3	≈ 37%	Typical vehicle suspension range
0.5	≈ 16%	Fast but bouncy
0.6	≈ 9.5%	Common servo target
0.707	≈ 4.3%	Butterworth pole angle, classic sweet spot
0.8	≈ 1.5%	Well damped, slower rise
1.0	0%	Critically damped

Worked example: designing a positioning stage

Suppose a linear positioning stage must move to a commanded position, overshoot no more than 5%, and settle to within 2% of the target in 0.5 s. Design the target closed-loop second-order dynamics.

1. Pick ζ from the overshoot spec. Invert the overshoot formula: ζ = −ln(0.05)/√(π² + ln²(0.05)). Here ln(0.05) = −2.996, so ζ = 2.996/√(9.87 + 8.98) = 2.996/4.34 ≈ 0.69. Round to ζ = 0.7, giving about 4.6% overshoot — safely under 5%.
2. Pick ωn from the settling spec. Require Ts = 4/(ζωn) ≤ 0.5 s. Solve ωn ≥ 4/(ζ·Ts) = 4/(0.7·0.5) = 11.4 rad/s. Choose ωn = 12 rad/s for margin.
3. Check the other metrics. ωd = ωn√(1−ζ²) = 12·√(1−0.49) = 12·0.714 = 8.57 rad/s. Peak time Tp = π/8.57 = 0.37 s. Rise time Tr = (π − arccos(0.7))/8.57 = (3.1416 − 0.795)/8.57 = 0.27 s.
4. Place the poles. s = −ζωn ± jωd = −8.4 ± j8.57. Any controller — a PD, a lead compensator, or a state-feedback gain — that drives the dominant closed-loop poles to this location meets the spec.

This four-line procedure is the everyday workflow of control design: translate time-domain specs into a target ζ and ωn, convert those into a desired pole location, then choose gains that put the poles there. Tools such as the root locus and Bode plot exist precisely to help you steer the dominant poles to that spot.

Common misconceptions and failure modes

• "More damping is always safer." Past ζ=1 the response only gets slower, not steadier. Overdamped ≠ better.
• "Overshoot depends on how fast you drive it." No — overshoot is a function of ζ alone; ωn changes only timing, not the overshoot percentage.
• "ωn is the frequency I'll observe." The ringing frequency is ωd = ωn√(1−ζ²), always below ωn; only for ζ=0 do they coincide.
• "A nearby zero doesn't matter." A left-half-plane zero close to the poles increases overshoot and speeds the rise; the pole-only formulas underestimate it.
• "Any high-order system reduces to second order." Only if the non-dominant poles/zeros are roughly 5–10× farther left; otherwise the approximation breaks down.
• "Settling time scales with ωn." It scales with the product ζωn (the real pole part σ), so a high-ωn but very lightly damped system can settle slowly.