The colours of noise are technical descriptions, not marketing. White noise has equal power at every frequency, which sounds hissy and harsh because human hearing groups frequencies into octaves — and the top octave contains half the audible range, so half the energy lands there. Pink noise falls off at 3 dB per octave, giving equal power per octave, which is why it sounds like rain rather than static.
Brown noise falls faster still at 6 dB per octave, which is why it sounds like a distant waterfall and why people reach for it to mask low-frequency sound.
Why does equal power sound wrong?
Because the ear does not hear in linear frequency. The octave from 20 to 40 Hz spans 20 Hz; the octave from 10,000 to 20,000 spans 10,000 — five hundred times as much bandwidth for the same musical interval.
| Noise | Roll-off | Character | Common use |
|---|---|---|---|
| White | flat | Hissy, bright | Testing, tinnitus masking |
| Pink | −3 dB/octave | Balanced, like rain | Sleep, room tuning |
| Brown | −6 dB/octave | Deep, like a waterfall | Masking traffic and voices |
Pink noise is the one used to measure a room’s response for exactly this reason — equal power per octave means a flat measurement corresponds to a flat perceived balance.
Why generate rather than stream?
Because noise is trivially cheap to produce and expensive to transmit. A few lines of arithmetic generates an unlimited, never-repeating stream in the page; streaming the same thing means downloading megabytes continuously and hearing an audible loop point when the file restarts.
The loop is the tell. Anything that repeats every few minutes is a recording, and once the repeat is noticed it cannot be un-noticed — which defeats the point of using it to fall asleep to.
What is the tone generator actually for?
Three jobs, and the third comes with a caution. A440 is the tuning reference for instruments. Sweeping upward is a rough way to find where your own hearing stops — most adults lose the top of the range progressively, and by middle age 15 kHz is often inaudible.
The third is testing equipment. A square wave is the honest test of a speaker or a signal chain, because it contains a fundamental plus every odd harmonic — anything that colours the sound shows up in how badly the corners round off.
Two cautions are worth taking seriously. Sustained pure tones damage hearing at volumes that do not feel dangerous, because a single frequency lacks the cues that make broadband sound feel loud. And low frequencies at high volume can damage speakers that are not designed to reproduce them, particularly small drivers asked for 30 Hz.
How loud is safe?
Quieter than most people assume for sustained exposure. The decibel article covers why the scale is logarithmic; the consequence here is that a small increase in the number is a large increase in energy.
The rule that generalises: if you would raise your voice to be heard over it, it is loud enough that duration starts to matter. For noise used to sleep or concentrate, the useful volume is the lowest one that masks what you are trying not to hear.
Masking works by raising the floor rather than by cancelling anything, so the required level is set by the sound you are covering. A quiet room needs very little.
Frequency weighting is the other reason a level meter and your ear disagree. Sound level meters apply an A-weighting curve that de-emphasises low frequencies to approximate human sensitivity, so a reading of 60 dBA on brown noise represents far more acoustic energy than 60 dBA on white noise. The number is comparable between measurements and not between spectra.
What about the click test?
It measures technique more than speed past a point. Around six clicks per second is ordinary; beyond about seven, people are not clicking faster but using a different motion — tensing the arm to induce a tremor, or dragging a finger across the button.
That makes the number comparable within a technique and meaningless across them, in the same way a reaction time figure is comparable only against your own previous attempts. Both are measuring a chain that includes the display, the input device and the operating system as well as you.
Questions people ask
Does noise actually help concentration? The evidence is mixed and individual. What it reliably does is mask intermittent sound, and intermittent sound is far more disruptive than steady sound at the same level.
Is brown noise the same as red noise? Yes — the name comes from Brownian motion rather than the colour, and both terms describe the same 6 dB per octave slope.
Can a tone generator damage headphones? A sustained low-frequency tone at high volume can, particularly on small drivers. Start quiet and raise it slowly.
Why does A440 vary? It was only standardised internationally in 1955, and baroque ensembles commonly play at 415 Hz. The semitone arithmetic covers what a reference pitch actually fixes.
Is there a grey noise? Yes — noise shaped to an equal-loudness contour so it is perceived as flat rather than measured as flat. It is the psychoacoustic version of the same idea and it is far less common.
The colours are slopes, and the slope is what you are choosing between. The white noise generator produces all three in the page rather than streaming them, the tone generator covers the single-frequency cases, the bars to seconds calculator handles musical time, and the click speed test is the other measurement whose number depends on how it was produced.