Sound is a pressure wave. If your headphones play a wave that is inverted relative to the noise (roughly 180° out of phase), the pressures add toward zero at your ear.
In practice, the headphone must predict what the noise will be by the time the anti-noise reaches your eardrum. That introduces two big limits:
Microphone capture, processing, and speaker output take time. If the anti-noise arrives late, it won’t line up and cancellation weakens—or can even make certain bands louder.
Low-frequency sounds (engine rumble, HVAC) change more slowly, so small timing errors still leave decent cancellation. High-frequency sounds (clinking dishes, many speech components) change fast; tiny delays or fit differences cause big phase mismatches.
A poor seal changes the acoustics of the earcup/ear canal, making the system’s prediction wrong. Leaks also reduce passive isolation, which is your main defense against higher frequencies.
ANC is strongest on steady, low-frequency noise; it’s weakest on sudden, high-frequency, and highly variable sounds.
Why do noise-canceling headphones typically reduce airplane engine hum more than nearby conversation?
ANC shines when the noise is predictable and slow-changing—exactly what low-frequency engine hum looks like—so the anti-noise can line up in time and phase. It’s easy to think loudness is the issue, but ANC is more about timing/phase than raw volume. Distance doesn’t automatically make sound “easier” to cancel at the ear. Microphones can measure conversation just fine; the problem is that speech is complex and changes quickly, so small delays and fit variations reduce cancellation.