IMMERSEDEAR LAB

DIGITAL AUDIO LAB / AN INTERACTIVE EXHIBIT

How often?
How precise?

DIGITAL AUDIO / ANOTHER EXPERIMENT

Nyquist Lab

Push a tone past the frequency ceiling. See and hear it fold back.

Open Nyquist Lab →

PERCEPTION / LISTENING EXPERIMENT

Equal-Loudness Lab

Same EQ, different impression. Explore listening level and draw your own tone-matching curve.

Open Equal-Loudness Lab →

WAV / MP3 · up to 30 MB / 2 minutes. This lab uses the first 15 seconds. Your original stays untouched; examples are created locally in your browser.

Use the practice voice or drop in your own recording.

1 / SEE IT

Sample rate: how often?

Continuous motion8 snapshots / second

Fewer snapshots → less information about change over time.

This analogy explains when measurements happen. Low-rate audio loses bandwidth; it does not simply become choppy like low-frame-rate video. Your display limits the visible frame rate.

2 ms · same 1 kHz signal · 8,000 samples/second

Dots are measurements, not stair steps. A correctly reconstructed, band-limited signal is continuous. Extra points are not automatically an audible improvement.

2 / HEAR IT

Same voice. Different bandwidth.

Load your voice or the practice VO above to listen.

Versions become available once your audio is ready.

Choose a setting to hear it immediately on this recording. Switch between Original and the selected version at the same position, without taking the quiz.

Start with 8 kHz, then try 48 kHz. Listen for missing high-frequency detail and a narrower, telephone-like quality.

Filtering prevents most aliasing before downsampling. This demonstrates bandwidth loss—not intentionally broken resampling. Pitch and duration stay the same.

4 / UNDERSTAND IT

More capacity is not a promise of better sound.

Higher sample rates allow a wider frequency band. Once the useful content is represented, extra samples need not produce an audible improvement.

The browser decoded this source at 48 kHz. All comparisons return to that playback rate. A 96 kHz version here is an upsampled comparison, not a new 96 kHz recording; it cannot restore missing high frequencies.

Technical view

8,000 samples per second · Nyquist limit: 4 kHz. Practical filters leave a transition band below that limit. Content above it must be removed before reducing the sample rate to avoid folding into lower frequencies (aliasing).

Examples use a Blackman-windowed sinc resampler with anti-alias filtering and band-limited reconstruction. The plotted sine is an illustration, not analysis of your voice.

Explore sampling with Xiph’s Digital Show & Tell