White vs. Pink vs. Brown Noise: Key Differences

Three Slopes, One Axis

Set the three side by side and one quantity separates them: the slope of the spectrum. White noise is flat, pink falls three decibels per octave, and brown falls six. Everything else people say about these sounds — that white is harsh, that pink is balanced, that brown is soothing — follows from where the energy ends up once the slope is fixed.

A slope describes how power spectral density changes with frequency. Flat means a one-hertz band at 100 Hz carries the same power as a one-hertz band at 10,000 Hz. Three decibels per octave means the power drops by half for every doubling of frequency; six means it drops faster still. The colour names borrow from optics, where white light carries every visible wavelength at roughly equal intensity, and they are a loose analogy rather than a definition. The names of the shapes come from different places — pink from the optical analogy, brown from Robert Brown and the Brownian motion whose mathematics it shares — but both describe a straight line on a logarithmic frequency axis.

White Noise: The Flat Baseline

White noise is the reference from which the other colours are derived, and the flat spectrum is what makes it useful. Feed it through a filter, compare the input spectrum to the output spectrum, and the filter's frequency response falls out directly. Play it through a loudspeaker, record it with a measurement microphone, and deviations from flat reveal the room's resonances and nulls. Band-limited is the practical caveat: a mathematically true white-noise signal would have infinite bandwidth and infinite power, so a digital signal runs from DC up to the Nyquist limit — 22.05 kHz at a 44.1 kHz sample rate.

Flat per hertz does not mean flat to the ear, and this is the single most common misunderstanding about the colour. Human pitch perception is logarithmic, grouping frequencies into octaves, and each octave spans twice the hertz of the one below it. The octave from 10 kHz to 20 kHz contains ten thousand hertz; the octave from 250 Hz to 500 Hz contains two hundred and fifty. Equal power per hertz therefore puts forty times more energy in the top octave than in that lower one. The hiss listeners describe is not a defect in the signal. It is the flat-per-hertz spectrum meeting a logarithmic ear.

Pink Noise: Equal Energy per Octave

Pink noise applies a minus-three-decibel-per-octave slope to white noise, and three decibels per octave is exactly the amount needed to cancel the doubling described above. The result is equal power in every octave band, which is why pink noise reads as flat on a one-third-octave analyser while white noise shows a rising trend on the same instrument. That correspondence is the reason pink noise, not white, is the usual signal for checking a sound system against an octave-band target.

Generating it digitally takes more care than generating white noise. Cascading first-order low-pass filters with staggered cutoffs approximates the slope. The Voss-McCartney algorithm takes a different route, summing white-noise sources updated at successively halved rates so that only one generator changes per sample, which holds the computational cost constant per sample rather than letting it grow with the number of octaves covered.

Line chart with frequency from 20 Hz to 20 kHz on a logarithmic axis and relative level from 0 to -60 decibels. Three straight lines start together at 0 decibels at 20 Hz. White noise stays flat at 0 dB per octave. Pink noise falls at 3 decibels per octave, reaching about -30 decibels at 20 kHz. Brown noise falls at 6 decibels per octave, reaching about -60 decibels at 20 kHz. 0 -20 -40 -60 20 100 1k 10k Relative level (dB) Frequency (Hz, logarithmic) White, 0 dB/oct Pink, -3 dB/oct Brown, -6 dB/oct
White, pink and brown noise on one logarithmic frequency axis, all three normalised to the same level at 20 Hz. The slope is the definition; the spread at the right-hand edge is what the ear hears.

Read the figure from the left edge, where all three signals are normalised to the same level, and the consequence of the slope is immediate. By 20 kHz — just under ten octaves up — pink has fallen about thirty decibels and brown about sixty. The vertical distance between the lines at the right-hand edge is the audible difference between the three colours, expressed as a quantity rather than an adjective.

Brown Noise: Integrated White

Brown noise falls at minus six decibels per octave, and the signal is generated by integrating white noise: each sample is the running sum of the random values before it. That integration is the same random-walk process that describes Brownian motion, which is where the name comes from. Doubling the rolloff relative to pink noise concentrates the energy heavily at the bottom of the spectrum. Between 100 Hz and 10 kHz — six and two-thirds octaves — the level falls roughly forty decibels, so the treble content is close to inaudible against the low end.

Two problems follow from the integration. Because each sample depends on the last, the signal can wander away from zero and eventually clip, which a gentle high-pass filter near 10 Hz corrects without touching the audible range. And because integration amplifies low-frequency energy so strongly, the output level needs normalising before anything downstream sees it, or the bass alone will drive the signal into distortion.

The practical consequence of a minus-six-decibel slope is that playback equipment matters more for brown noise than for the other two colours. Most of the signal's energy sits below 200 Hz, which is where small speakers and earbuds roll off steeply, so the same file can sound full on one system and thin and quiet on another. Nothing is wrong with the signal in that case: the part of the spectrum carrying the energy is simply not being reproduced.

Comparing the Three Spectra Side by Side

White vs. pink vs. brown noise at a glance
PropertyWhite noisePink noiseBrown noise
Spectral slope0 dB/octave (flat)-3 dB/octave-6 dB/octave
Energy emphasisEqual power per hertz (bright, treble-heavy)Equal power per octave (balanced)Low-frequency heavy (deep)
How it soundsBright, sharp, hissyBalanced, warm, naturalDeep, rumbly, soothing
What it masksBroadband, with enough treble energy to cover clicks, latches and other transientsBroadband, with proportionally more power in the 300 Hz to 3 kHz speech range than whiteLow-frequency content; little energy left above a few hundred hertz to cover anything higher
Measurement useSpectral analysis: filter and loudspeaker responseOctave-band work: system checks against an octave-band targetNot used as a test signal
AnalogyWhite light (all frequencies equal)Reddish balance of low and highDeep red / infrared end

The table stops at the spectrum on purpose. Where a colour’s energy sits is a different question from what listening to it does, and the second one needs evidence this article does not carry. For that, the brown noise and white noise safety articles on this site cite the sleep research directly, including how the systematic reviews grade it.

The spectrum is not the only thing the slope changes. Successive samples of white noise are uncorrelated — each one is independent of the last. Pink noise carries short-term correlations, which give it a slightly smoother texture. Brown noise is strongly correlated by construction, since every sample contains the whole history before it, and on a waveform display this shows as slow, wandering undulations rather than the dense spiky band that white noise produces. Two signals can differ audibly in texture while sharing an overall level, and this is where that difference comes from.

Choosing a Colour

For measurement, the use case picks the colour and there is little room for preference. Flat per hertz suits spectral analysis, so white noise is the default for characterising filters and loudspeakers. Octave-band work calls for pink noise, because the analyser and the signal share the same assumption about what flat means. Mixing the two is a common way to misread a measurement: a system that looks like it has a rising treble response under white noise on an octave-band analyser may be perfectly flat, with the slope belonging to the test signal rather than the system under test.

For listening, the ambient noise already in the room matters more than any ranking of the three colours. A room with heavy low-frequency content from traffic or ventilation is a poor candidate for brown noise, which reinforces exactly the rumble that is already there; white noise or a shaped spectrum complements it better. A quiet room where the intrusions are high-frequency transients — keyboard clicks, notifications, a door latch — is the opposite case, and the extra treble energy in white noise is doing useful work there.

Comfort over a long session is the constraint most listeners underestimate. The treble emphasis that makes white noise effective against transients is also what makes it fatiguing after an hour, which is why pink and brown are the more common choices for sleep and for sessions measured in hours rather than minutes. Thirty seconds of each through headphones in a quiet room is usually enough to settle the question, because the spectral differences here are large rather than subtle.

References

Frequently Asked Questions

Which noise color is best for blocking out conversations?

Pink noise is usually more effective than white for masking speech, because equal energy per octave puts proportionally more of its power in the speech range of roughly 300 Hz to 3 kHz instead of spending it above 5 kHz where speech has little content. A spectrum shaped specifically for the speech bands does better still.

Why is brown noise called 'brown' if it is not related to the color?

The name comes from Robert Brown, the botanist who first described Brownian motion. Brown noise is generated by integrating white noise, which is mathematically equivalent to the random walk process that describes Brownian motion.

Can I create pink noise by simply turning down the treble on white noise?

A simple treble reduction approximates the effect, but true pink noise requires a precise minus-three-decibel-per-octave filter slope across the entire spectrum. A standard tone control does not provide this exact slope.

Is there a noise color between pink and brown?

Yes. Any slope between minus three and minus six decibels per octave is achievable, and such intermediate shapes are sometimes called red noise, though the terminology is inconsistent and "red" is also used as a synonym for brown. Unlike white, pink and brown, these in-between slopes have no standard definition.

WhiteNoise.top

WhiteNoise.top builds free browser-based noise generators, mixers and breathing timers. Our articles explain the acoustics behind those tools and cite published research where it exists.