The poles and zeros of a transfer function are the roots of its denominator and numerator polynomials, respectively. They’re the geometric handles for understanding an LTI system’s dynamics in the s-plane.

Image: Pole–zero plot in the s-plane, CC BY-SA 4.0 — poles marked ×, zeros marked ○.

For with both polynomials in :

  • Zeros: values of where — the roots of .
  • Poles: values of where — the roots of .

Why pole locations matter

The poles of are the eigenvalues of the system’s differential equation — the rates at which the system decays in the absence of input. So they encode every time-domain feature of the impulse response:

  • Real part of pole : decay (or growth) rate. Negative → decaying transient; positive → growing (unstable).
  • Imaginary part of pole : oscillation frequency. Nonzero imaginary part → oscillation.
  • Distance from origin : natural frequency , the speed of the system’s response.
  • Repeated poles of order at : bring in the full set of polynomial-times-exponential terms . The highest-order term dominates at large , but all the lower-order terms appear in the partial-fraction expansion and contribute to the early-time response.

A first-order pole at corresponds to a decaying exponential with time constant . A pair of complex-conjugate poles at corresponds to a damped sinusoid .

Stability from poles

For a causal LTI system, BIBO stability requires all poles in the open left half-plane (negative real parts). Poles in the right half-plane → unstable (growing). Simple poles on the imaginary axis → marginally stable (sustained oscillation, neither growing nor decaying). Repeated poles on the imaginary axis → unstable (growing oscillation).

This pole-location criterion is one of the most-used tools in classical control theory and filter analysis.

What zeros do

Zeros don’t determine stability — they don’t appear in the system’s natural response. But they shape the frequency response: at a zero on the imaginary axis , the system completely blocks the frequency . This is how notch filters kill specific frequencies (60 Hz hum, for instance).

In the Bode plot, a zero adds dB/decade to the magnitude slope past its corner frequency and adds to the phase; a pole adds dB/decade and . Zeros and poles “cancel” each other when they’re close, and the spacing between them shapes the rolloff.

Pole-zero plot

A pole-zero plot is just a sketch of the s-plane with poles marked as × and zeros as ○. From the plot you can read off the system’s behavior at a glance:

  • All בs in the left half-plane → stable.
  • Any × on or right of the imaginary axis → unstable or marginally stable.
  • ○ at → DC zero (signal is “AC-coupled” through this system).
  • ○ near a ×: zero-pole cancellation; the corresponding mode is mostly suppressed.

Filter design language

Filter design becomes the engineering art of placing poles and zeros on the s-plane to get a desired frequency response. Different filter families (Butterworth, Chebyshev, Elliptic) correspond to different pole/zero placement strategies. The trade-offs are between passband flatness, transition bandwidth, stopband attenuation, and order. See Filter (signal processing).