Brown noise vs white noise: how to actually choose

Short answer: white noise is flat, with equal energy at every frequency, so it sounds like hiss or untuned radio static. Brown noise rolls off steeply towards the top, losing roughly 6 dB per octave as you go up, so it sounds like a low roar or distant waterfall. Choose brown when the thing disturbing you is low and rumbling (traffic, a boiler, footsteps upstairs, bass through a wall). Choose white or pink when the thing disturbing you is bright and sharp (voices, a keyboard, cutlery, a door catch). Then check your speaker can actually play what you picked, because most phones cannot really produce brown noise at all.

That last point is the one nobody mentions, and I will come back to it.

The difference in one paragraph

All of these "colours" describe the same thing: how energy is distributed across frequency.

• White noise has equal power per hertz. Because each octave upwards contains twice as many hertz as the one below, white noise puts most of its energy in the top octaves. To the ear it is bright, hissy, slightly harsh.
• Pink noise falls by about 3 dB per octave, which works out as equal power per octave. It sounds like steady rain or a big waterfall heard from a hundred metres away. Most people find it the most "natural" of the three.
• Brown noise (also called Brownian or red noise) falls by about 6 dB per octave. It is dominated by the bottom two or three octaves and sounds like a jet cabin, a motorway from indoors, or a very heavy downpour.

There is nothing mystical in any of them. They are the same random signal filtered differently. Which means the question "brown noise vs white noise" is really a question about filters, and the correct filter depends entirely on what you are trying to cover up.

Match the noise to the intruder, not to your mood

Masking works by frequency overlap. A masker only hides sounds that sit in roughly the same part of the spectrum as itself. Brown noise, with almost all of its energy below 500 Hz, is very good at burying a lorry idling outside and close to useless against a conversation in the next room.

This is where I want to name something, because it comes up constantly and I have never seen it stated plainly.

The consonant window: speech intelligibility lives mostly between about 2 and 6 kHz. That is where fricatives and plosives (the s, f, t, k sounds) carry the information your brain uses to decode words. Brown noise has effectively nothing in that band. You can layer brown noise over a podcast in the next room and still follow the podcast.

Which is why the standard advice "brown noise is better for focus" is half wrong. It is better for focus if what breaks your focus is rumble. If what breaks your focus is other people's voices, and for most open offices and most flats it is, then you need energy in the consonant window, and that means pink or white. Brown will make the room feel warmer and the voices no less intelligible.

A quick mapping I use:

• Traffic, plumbing, HVAC, bass from a neighbour's TV, footfall: brown.
• Speech, television dialogue, phone calls, birdsong, keyboards, dishes: pink or white.
• Mixed household chaos, which is most homes: brown for the body plus a thin layer of pink or white on top. This is the combination I reach for most, and it is why the noise in Tysha is mixed live rather than shipped as fixed files. You cannot pick the right filter in advance for someone else's flat.

The small-speaker tax

Here is the finding from building this that changed how I think about the whole question.

The small-speaker tax: a phone speaker, a laptop speaker or a pocket Bluetooth speaker cannot usefully reproduce the low frequencies that define brown noise. What comes out is the thin remainder above the driver's roll-off, which measures and sounds much closer to pink. You selected brown; the room heard something else.

You can test this yourself in about two minutes. Play brown noise through your phone's built-in speaker, record it with a second phone from a metre away, and look at the spectrum in any free analyser. Then do the same with headphones on a dummy setup or a real speaker with a woofer. The recorded curves are not the same shape. The physics is unremarkable: a driver a few millimetres across moving almost no air cannot produce meaningful output at 80 Hz.

The practical consequences are worth spelling out:

• If you are playing sound from a phone lying on a bedside table, the brown-versus-white decision matters less than you think, because the bottom of brown is being discarded before it reaches you. Pick by how the top end sounds, since that is all you are getting.
• Brown noise pays off on hardware that can actually move air: headphones, a speaker with a real woofer, a decent soundbar, a car.
• Turning the volume up to compensate does not restore the missing bass. It makes the surviving midrange louder, which is the opposite of what you wanted brown for.
• Placement partly rescues it. A phone on a hard surface near a wall or in a corner gains some low-end reinforcement. A phone face-up on a duvet loses more.

I have watched people conclude "brown noise does nothing for me" when what actually happened is that they never heard brown noise.

Why brown feels gentler, and how much of that is real

Human hearing is most sensitive somewhere around 2 to 5 kHz. White noise dumps a lot of energy exactly there. So if you compare brown and white at the same slider position, or the same nominal signal level, white will sound substantially louder as well as brighter. Much of the reported preference for brown is a loudness comparison in disguise.

If you want an honest comparison, match them by ear before you judge. Play white, set it to a level you would tolerate for an hour, then switch to brown and adjust until it feels equally present. Now compare. In my experience some of the preference survives this test and some of it evaporates. The part that survives is real: sustained energy in your most sensitive band is genuinely more fatiguing, which is why long sessions of white noise can leave you feeling slightly wrung out in a way that brown does not.

One consequence for anyone building or choosing a sound app: a linear volume slider makes this comparison almost impossible. At the quiet end, one pixel of travel is a large perceptual jump. We map volume perceptually rather than linearly for exactly this reason, so the bottom third of the slider has usable resolution. It sounds like a triviality. At 3am with one hand free it is the difference between finding the right level and giving up.

Brown noise vs white noise for sleep

For sleeping adults, brown tends to win on comfort and white tends to win on coverage, and which you need depends on your building. If you are woken by a specific recurring sound, identify its pitch before choosing. Hum, rumble and thud are low. Latches, taps, voices and alarms are not.

Two things I would flag from listening to a lot of these files at unsociable hours:

• Loop seams. A great deal of sleep noise is a short rendered file played on repeat. Even a well-crossfaded loop has a periodicity, and a brain lying awake in the dark is extremely good at finding periodicity. Once you have noticed the seam you cannot unnotice it. Noise generated continuously has no loop point, and this matters more to me than the exact spectral slope.
• Abrupt starts and stops. Sound arriving instantly at full level is startling, and silence arriving instantly is more startling still. Fades of a couple of seconds at both ends cost nothing and change the experience completely.

I am not going to tell you anything about what noise does or does not do for a baby's sleep. That is not my field and I do not know. What I can tell you is what the sound itself is doing, which is the part I can measure.

Brown noise vs white noise for focus and tinnitus

For focus in a room with other people in it, start with pink. It covers the consonant window without the harshness of full white, and it is the least tiring of the three over a four-hour stretch. If the voices are still intelligible, move towards white rather than turning pink up, since more level in the wrong band mostly just gets louder.

For tinnitus masking, the honest answer is that it depends on your tinnitus and this is a question for an audiologist, not a blog. Perceived pitch varies enormously between people, and a masker that sits nowhere near your tone will not help however pleasant it sounds. If you experiment, the useful variable is not the colour name but whether the sound has energy at the pitch you hear.

A 60-second way to decide

1. Sit in the room at the time of day that is actually a problem. Listen for thirty seconds and name the intruding sound. Is it low and continuous, or bright and intermittent?
2. Pick brown for low and continuous, pink or white for bright and intermittent.
3. Set the level with the intruding sound present, not in silence. Noise set in a quiet room is almost always too quiet later.
4. Check your playback device. If it is a phone speaker and you chose brown, either accept you are hearing something closer to pink, or use headphones or a speaker with a woofer.
5. Leave it running for a full session before deciding. The colour that sounds best in a ten-second A/B is frequently not the one you can tolerate for six hours.

My own rule for level, for what it is worth as a personal habit rather than advice: if I would have to raise my voice to talk over it, it is too loud, and I turn it down.

What actually separates good noise from bad noise

Having spent an unreasonable amount of time on this, my ranking of what matters, most important first:

1. Does it play at all, instantly, offline? A buffering spinner at 3am is worse than no sound.
2. Does it loop? Continuous generation beats a clean file with a seam.
3. Can you set the level precisely at the quiet end?
4. Does it fade in and out?
5. Is the spectrum right for your room?

The brown-versus-white question sits at number five. It is a real question and worth getting right, but four things above it will ruin the experience first. Get those in order and the colour becomes a preference you can adjust in ten seconds rather than a decision you have to research.