Skip to content
Tutorials Center
My learning

Topics

Lesson 15 of 37 ยท Stability

Stability and Pole Locations

Learn what stability means for control systems, BIBO and marginal stability, how pole locations decide it, and test stability with NumPy roots and simulation.

A controller that makes a system unstable is worse than no controller. An unstable loop is one where small disturbances grow instead of dying out: a microphone howling, a Segway toppling, a power grid oscillating until it trips. Stability is therefore the first requirement of every control design, and the first thing to check. This lesson defines it carefully and shows how pole locations decide it.

What stability means

There are several related definitions for a linear time-invariant (LTI) system.

  • Asymptotic stability: with zero input, the state returns to zero from any initial condition.
  • BIBO stability (bounded-input, bounded-output): every bounded input produces a bounded output.
  • Marginal stability: the output neither decays nor grows for some inputs, for example a pure oscillation.
  • Instability: some modes grow without bound.

For LTI systems with no hidden (cancelled) unstable modes, asymptotic and BIBO stability are equivalent, and both are decided by the poles.

The pole-location rule

Each pole p of G(s) contributes a mode e^(p t) to the response. Writing p = sigma + j w, the magnitude of that mode is e^(sigma t).

Pole location Mode behaviour Verdict
sigma < 0 (left half-plane) decays stable
sigma > 0 (right half-plane) grows unstable
sigma = 0, simple (distinct) constant or steady oscillation marginally stable
sigma = 0, repeated grows like t unstable

So a system is stable if and only if all its poles lie strictly in the left half of the complex plane. Equivalently, the roots of the characteristic polynomial all have negative real parts.

Stability and the s-plane

Left half-plane Modes decay. Stable.
Imaginary axis Steady oscillation. Marginal.
Right half-plane Modes grow. Unstable.

Every pole must be strictly to the left.

1 / 2

Testing stability with roots

For a transfer function, compute the roots of the denominator and inspect their real parts.

Python Classify systems from their poles

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The first two systems are stable; 1/(s^2 + 4) has poles at +/- 2j (marginal); the others have a pole with positive real part. The cubic s^3 + s^2 + s + 6 is a good example: from its coefficients alone it is hard to tell, but its roots reveal two in the right half-plane. In the next lesson we will test stability from the coefficients without solving for roots.

Marginal stability and resonance

A marginally stable system looks harmless but is dangerous: its response to a bounded input can be unbounded. For G(s) = 1/(s^2 + 1) (poles at +/- j), a sinusoidal input at exactly 1 rad/s drives it at resonance and the amplitude grows without limit. Therefore marginal stability is not BIBO stable.

Python Resonance of a marginally stable system

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The peak amplitude grows linearly with time (about t/2), even though the input never exceeds 1. This is why design requires poles strictly inside the left half-plane, with some margin.

Stability of the closed loop

For a feedback loop, the closed-loop poles are the roots of the characteristic equation 1 + L(s) = 0, that is D_L(s) + N_L(s) = 0 for L = N_L/D_L. A stable plant can have an unstable closed loop, and an unstable plant can be stabilised by feedback.

Consider the plant G(s) = 1/(s (s + 1)^2) (an integrator with a double pole at -1) with proportional gain K. The closed-loop characteristic polynomial is s^3 + 2 s^2 + s + K. Scan K:

Python Closed-loop poles as the gain increases

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

For K < 2 all poles are in the left half-plane. At K = 2 the polynomial s^3 + 2s^2 + s + 2 = (s + 2)(s^2 + 1) has a pair on the imaginary axis (marginal), and for larger K it is unstable. The loop has a critical gain K = 2: the exact boundary of stability, found analytically with the Routh criterion in the next lesson.

Absolute versus relative stability

Being stable is yes or no, but a design that is stable by a hair is useless since models are imperfect. Relative stability asks how far from the boundary the system is, measured by the distance of the poles from the imaginary axis (or later by gain and phase margins in Chapter 5).

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. A system has poles at -1, -3 and 0.2. Is it stable?

    All poles must be strictly in the left half-plane; the pole at +0.2 gives a growing mode e^(0.2t).
  2. 2. G(s) = 1/(s^2 + 9) driven by sin(3t) produces

    Poles at +/- 3j mean the system is marginally stable; forcing at the natural frequency gives resonance, so a bounded input produces an unbounded output.
  3. 3. What determines closed-loop stability?

    The closed-loop poles are the roots of the characteristic equation 1 + L(s) = 0.
  4. 4. The closed-loop polynomial is s^3 + 2s^2 + s + K. For which K is it marginally stable?

    At K = 2 it factors as (s + 2)(s^2 + 1), giving poles at +/- j on the imaginary axis.

Summary

  • An LTI system is stable if and only if all its poles are strictly in the left half-plane.
  • Poles on the imaginary axis give marginal stability, which is not BIBO stable because resonance occurs.
  • Closed-loop stability depends on the roots of 1 + L(s) = 0 and changes with gain.
  • np.roots gives a quick stability check; relative stability measures how far you are from the boundary.

Next: The Routh-Hurwitz Criterion