White, Pink and Brown Noise: What the Colours Mean

Three buttons on a noise generator, three names borrowed from light: white, pink and brown. It reads like branding, but the names track a genuine physical difference in how each signal spreads its energy across the spectrum, and that difference is why one sounds like static, one sounds like rain, and one sounds like the inside of an aircraft cabin. None of the three is "just noise" — each is a specific, definable shape, and the shape is what a speaker-measurement engineer, a sleep-app developer and a bored commuter are each actually reaching for when they pick one over the others.

Equal energy per hertz, and why that isn't neutral

White noise is defined by a flat power spectral density. Rational Acoustics' own guide to the three colours, written for engineers using its Smaart measurement software, puts the definition plainly: white noise "contains equal energy across all frequencies" and has "equal power in each bandwidth" — the band from 1,000 to 1,001 Hz carries exactly as much power as the band from 15,000 to 15,001 Hz. On an instrument that reads a fixed number of hertz at a time, that really is neutral. Hearing does not read frequency in fixed hertz-wide slices; it groups pitch by octave, a doubling of frequency, and an octave's bandwidth grows every time it moves up. The octave from 1,250 Hz to 2,500 Hz spans 1,250 Hz of bandwidth; the octave from 10,000 Hz to 20,000 Hz, three octaves higher, spans 10,000 Hz — eight times as much. Equal energy per hertz means that top octave carries eight times the power of the one three octaves below it, and an ear that groups by octave hears that as a signal loaded toward the top: thin, hissy, closer to an untuned radio than to anything felt as balanced.

Equal energy per octave: pink noise's −3 dB slope

Pink noise is built to cancel exactly that imbalance. Rather than holding power flat per hertz, its power falls as frequency rises, at the specific rate that keeps every octave's total energy equal to every other octave's. The same guide states it directly: pink noise "has equal energy per octave," produced by a spectrum whose "intensity level is reduced by 3 dB/octave." That −3 dB-per-octave tilt is the whole trick: it exactly cancels the doubling bandwidth of each higher octave, so no octave dominates the way the top of white noise's spectrum does. That is also why pink, not white, is the reference signal audio engineers reach for. The same source notes that pink noise "is used to measure the frequency response of a room or a system," while white noise, despite sometimes serving as "a reference tone for Transfer Function measurements," is described in the same guide as "not ideal" for that job on the logarithmic, octave-based frequency plots audio measurement actually uses.

Brown noise: a steeper slope, and a name that isn't about colour

Brown noise, also called red noise, falls faster still. Where pink loses 3 dB per octave, brown loses twice that: an EBSCO Research Starters entry on Brownian noise gives it as "a −6dB/oct decrease in energy — or a power spectrum proportional to 1/frequency²." At that slope the bass dominates so heavily that most of the energy above a few hundred hertz has essentially vanished, which is why brown noise reads as a low rumble rather than a hiss or a hush. The name is a common trap: it sounds like a third colour swatch alongside white and pink, but the same EBSCO entry is explicit that it is not — brown noise "takes its name from Scottish botanist Robert Brown," whose 1827 observation of pollen jittering erratically in water gave Brownian motion its name, and the noise is called Brownian "because its signal changes follow a random-walk pattern," the same statistical process. A separate account on Wikipedia's Brownian noise entry corroborates the mechanism directly: brown noise "can be produced by integrating white noise" — literally summing a white-noise signal over time, the way position is the integral of a random walk's steps. "Brown" is a coincidence of English (the colour and the scientist's surname happen to be homophones once genericised); "red," the alternative name, is the one that is actually a colour analogy, chosen because red light sits at the low-frequency end of the visible spectrum the way brown noise's energy sits at the low-frequency end of the audible one.

What this site measured

The three colours this site's own noise generator plays were measured for spectral tilt using a first-difference energy ratio — comparing the energy in a signal to the energy in the difference between each sample and the one before it, a simple roughness measure that sits near 2.0 for an uncorrelated sequence and falls as a spectrum tilts toward low frequencies. White measured 2.008, matching the 2.0 expected of uncorrelated white noise almost exactly; pink measured 0.353; brown measured 0.039. The three fall in the order the −3 and −6 dB slopes predict, each roughly an order of magnitude apart. All three play at the same −20 dBFS starting level, and stay level-matched however you switch between them, because each buffer is peak-normalised before playback rather than played at whatever raw amplitude its algorithm happens to produce.

What each is actually used for

Pick the colour for the job, not the mood. Pink is the standard for level-matching a pair of speakers and for measuring a room or system's frequency response, precisely because its per-octave flatness lines up with how both hearing and octave-based measurement plots divide the spectrum — a broadband, steady signal that treats every octave the same way a swept tone cannot. White works where flatness-per-hertz is actually what is being asked for, or as a quick, everything-at-once check that something is producing sound across the full range; it is a poor choice for probing a system on the log-frequency plots measurement software draws, for the same reason it sounds thin rather than neutral to begin with. Brown and red trade measurement precision for comfort: their thin high end makes them the more pleasant of the three to sit through for a long stretch, but that same thin high end is exactly why they are a worse tool for finding fine detail or timing information across the full spectrum than a flatter signal is.

Sleep, tinnitus and focus: what the evidence actually shows

Search for any of these colours and you will find no shortage of claims about sleep, concentration and tinnitus relief, and most of that literature is thin, uncontrolled or commercial. One real data point exists for tinnitus: a small clinical study published in the American Journal of Translational Research followed 43 adults with tinnitus and hearing loss through pink-noise-based sound therapy delivered alongside their hearing aids, and found Tinnitus Handicap Inventory scores dropped significantly at one, three and six months. It is not a controlled comparison against white noise or against no treatment, and the authors say so themselves, writing that the sample size "may have affected the generalizability of this work" and that "further neurophysiologic research and extensive clinical observation are needed before this becomes a routine treatment." That is a genuine, cited result about one small trial — not a finding that pink noise treats tinnitus, and not something this site or its noise generator claims to do. If tinnitus is new, sudden or in one ear only, that is a reason to see a doctor, not a reason to reach for a browser tab.

This page measures neither your room nor your hearing; it describes what the three signals are and what they were built for, not what your speakers or ears will do with them. If you want the same "per hertz versus per octave" distinction applied to how digital audio itself is stored rather than to noise, sample rate and bit depth explained covers the adjacent ground, including where dithered quantisation noise behaves like a very quiet relative of white noise. Once a room is the variable rather than the signal, speaker placement and room acoustics covers what pink noise is actually being used to reveal when it is played for that purpose. And because noise carries far more total energy than a single tone at the same setting, how loud is too loud is worth reading before leaving any of the three colours running at a level you would not choose for a tone.