A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Gain margin (GM) is the factor by which open-loop gain can be increased before stability is lost at phase = −180°. Phase margin (PM) is the additional phase lag that can be tolerated before instability. These margins quantify robustness to unmodeled dynamics and parameter variations.
From the Nyquist criterion, you know that a closed-loop feedback system is unstable when the open-loop Nyquist contour encircles the −1 point in the complex plane. That's a geometric condition — powerful but abstract. Gain margin and phase margin translate this into two concrete numbers you can read directly from a Bode plot, and more importantly, into actionable design targets that tell you how much "safety buffer" your system has against the parameter variations and unmodeled dynamics that every real system carries.
Phase crossover frequency ω_pc is the frequency where the open-loop phase equals exactly −180°. At this frequency, a signal traversing the forward path and feedback path returns perfectly inverted — if the loop gain at this frequency equals or exceeds 1, the inverted signal reinforces itself rather than correcting the error, and the system oscillates or diverges. Gain margin (GM) is the reciprocal of the open-loop magnitude |G(jω_pc)H(jω_pc)| at that critical frequency. Expressed in decibels: GM(dB) = −20·log₁₀|L(jω_pc)|. A gain margin of 10 dB means you could multiply the open-loop gain by a factor of 3.16 before stability is lost. Larger GM means more robustness to gain variations — which arise from component tolerances, temperature drift, or nonlinear operating point shifts.
Gain crossover frequency ω_gc is the frequency where the open-loop magnitude equals exactly 1 (0 dB on a Bode magnitude plot). At this frequency, the loop has unity gain. Phase margin (PM) is the amount by which the open-loop phase exceeds −180° at ω_gc: PM = 180° + ∠L(jω_gc). If the phase at ω_gc is −145°, then PM = 35°. Phase margin quantifies how much additional phase lag the system can absorb before going unstable — this matters because unmodeled delays (sensor lag, actuator dynamics, digital sampling delay) all add phase lag. Every 1 ms of additional delay at frequency ω_gc contributes −ω_gc × 0.001 × (180°/π) degrees of phase lag, directly consuming phase margin.
A PM of 40–60° is the standard engineering target. This range corresponds to a second-order system with damping ratio ζ ≈ 0.4–0.7 — enough damping to prevent excessive overshoot in the step response while maintaining adequate bandwidth. PM below 20° produces a sluggish, oscillatory response with large overshoot. PM above 70° gives a smooth but slow response; the controller is being overly conservative. GM above 6 dB (factor of 2) is a common minimum rule of thumb. These are starting points, not universal rules — the appropriate margins depend on how well the model captures reality and what variation the system must tolerate in service. When you read a Bode plot, GM and PM are the first numbers to extract: they tell you immediately whether the design is viable or needs adjustment.