IMMERSEDEAR LAB
← Sample Rate / Bit Depth

DIGITAL AUDIO / NYQUIST LAB

Push past
the ceiling.

Every sample rate has a limit. What happens when a sound goes past it?

1 / SEE IT

Up to the wall. Back down again.

Sample rate8 kHz
Nyquist ceiling4 kHz
Represented frequency3 kHz
Nyquist wall · 4 kHzRepresented frequency04 kHz8 kHzIntended tone →

Below the ceiling: the represented frequency follows your tone.

2 / HEAR IT

Keep sliding up. Hear it come down.

Start with represented playback, then slide past 4 kHz.

At 8 kHz sampling, sweep from 3 kHz through 4 kHz to 6 kHz. The intended pitch rises; the represented pitch rises, reaches the wall, then falls.

This player synthesizes the calculated frequencies directly. It demonstrates unfiltered sampling; it does not change your device’s sample rate. Tones at 0 Hz or 18 kHz and above are muted. High tones may be inaudible on your hearing or playback system.

3 / TRY IT · Find the Ceiling

Where is the frequency ceiling for 32 kHz sampling?

Session: 0 / 0

4 / UNDERSTAND IT

You just met Nyquist.

Every sample rate has a frequency ceiling. That ceiling, called the Nyquist frequency, is half the sample rate. Above it, different frequencies produce indistinguishable samples. An unfiltered high tone can then appear as a lower tone: an alias.

Why half? A cycle needs more than two evenly spaced measurements for reliable representation of an arbitrary band-limited signal. A sample rate of 8,000 measurements per second therefore supports frequencies below 4,000 cycles per second. Exactly two samples per cycle is a special boundary, not a reliable target.

A higher sample rate raises the ceiling. Real recording systems use anti-aliasing filters to remove unwanted content above the limit before sampling. Filters need transition space, so usable bandwidth stops short of the theoretical ceiling.

Learn more / Technical view

Nyquist frequency = sample rate ÷ 2. In the first fold, alias frequency = sample rate − intended frequency. For frequencies beyond one sample-rate interval, the folding repeats.

The Nyquist–Shannon sampling theorem concerns band-limited signals. At the exact boundary, a sine may sample to zero depending on phase. At a multiple of the sample rate, the sampled signal can be a constant (DC); this exhibit mutes DC. Its graph shows frequency magnitude, not reconstructed phase or amplitude.

The audible comparison is an ideal frequency demonstration. Proper resampling in the main Digital Audio lesson filters before downsampling, preventing the fold-back intentionally demonstrated here.

Return to Sample Rate / Bit Depth →