Sound,
made visible.

What are you changing?

Frequency is the number of cycles per second, measured in hertz (Hz). A 500 Hz sine wave completes 500 cycles each second. In the lab, the carrier animation is slowed 1,000 times so you can see it move. Optional audio uses the displayed frequency.

Amplitude controls the signal’s strength. In a sound wave this relates to pressure variation, but perceived loudness also depends on frequency, your hearing and playback level. The lab’s amplitude percentage is a relative control, not a calibrated sound-pressure reading.

Harmonics are components at whole-number multiples of a fundamental frequency. The sixth harmonic rotates six times as quickly as the fundamental in the Fourier view. An octave is specifically a doubling of frequency, so harmonics are not all octaves.

An envelope changes amplitude over time: attack, decay, sustain and release. It shapes how a note begins, settles and ends. Envelope stages run at normal display speed while carrier motion is slowed, so the animation is a teaching model rather than a uniformly time-stretched recording. Audio is generated at the displayed frequencies with smoothing; harmonics above the playback sample-rate limit are omitted.

Why does the plate look different?

The plate is a finite-difference approximation of a clamped aluminium surface, using 64 bending modes on a 49-point grid per axis. Its outline has a 27.6 cm circumscribed diameter; thickness is 0.8 mm. Its shape changes the available modes of vibration. Illustrative particles drift towards lower-vibration regions. The calculation combines steady-state mean-square displacements from the separate harmonics; grain collisions, friction and transient ring-down are not solved. The displayed value is the maximum model RMS displacement across the surface. These are often called Chladni patterns, part of the study of cymatics. The UCLA Chladni plate demonstration shows how different mounting and excitation change the patterns.

A frequency alone cannot specify a universal pattern. A real plate’s dimensions, material, thickness, mounting, damping and drive position all matter. This is a truncated, discretised educational model, not a validated prediction of a particular real plate. Results at higher frequencies are limited by the grid and retained modes. Scaling a steady response with an envelope does not simulate transient resonance build-up or decay.

The dot matrix and propagation views are visual interpretations of the same signal. They help you compare changes, rather than measuring an actual room or sound field.

Rhythm is a relationship.

The rhythm study divides one shared cycle into different numbers of evenly spaced beats. Three beats trace a triangle; four trace a square. In a 3 : 4 polyrhythm, the layers meet at the start of each cycle. A 2 : 4 pattern is a simpler subdivision of the same pulse.

The numbers describe beats in a shared cycle, not time signatures. Tempo uses quarter-note beats per minute and each cycle lasts four quarter notes. You can layer up to twelve rhythms, change their note velocity and listen to each separately.

Link the rhythm study to switch from these constructed-wave models to four measurements of the actual digital audio output. The waveform, spectrum, spectrogram and amplitude envelope include the selected drum samples, recorded clips and original/rhythm blend. Manual wave settings remain stored until you unlink.

Inspect the actual audio.

Turn sound on in the linked mode to capture the mono fold-down of the stereo output, after listening volume and limiting, before your audio device. Stereo playback is unchanged. Opposite-phase left and right components can cancel in the mono analysis. This is a digital measurement, not acoustic pressure at your ears.

The waveform plots signed samples over a labelled millisecond window. Magnification changes only the display. The spectrum and spectrogram use the browser’s 4,096-sample Blackman-windowed FFT, with no additional smoothing. At 48 kHz, the window is about 85.3 ms and bins are about 11.7 Hz apart. Log-spaced display bands show their strongest bin, not integrated band energy. Nearby low-frequency bands may share a bin; more drawn bands do not mean more resolution. Spectrogram colour spans −90 to −15 FFT dB, saturating above that range.

FFT levels retain the browser’s transform scale. They are not calibrated single-tone amplitudes or dB SPL. The envelope view shows sample-window RMS and sample peak in dBFS, using amplitude 1 as the reference. It is not LUFS, true peak or a complete transient measurement. The six-second history samples windows about 20 times per second, or 10 in Eco mode, while visible. Missing capture periods are not fabricated.

Freeze views to keep listening while inspecting the captured data. Tap either history plot or use the inspection slider to select a snapshot shared by all four panels. These views use real audio time. A five-millisecond window magnifies time horizontally; it does not slow playback.