Brown noise for sleep: what actually matters at 3am

Brown noise for sleep does one useful thing: it raises the noise floor of your room so that the sounds which would otherwise pull you out of light sleep no longer stand out. It is masking, not sedation. Nothing in the sound makes you sleepy. What it does is make a fridge compressor, a partner turning over, a lorry two streets away or a neighbour's television less eventful — and events are what wake people. If you understand that, most of the practical questions answer themselves: how loud, for how long, on what speaker.

Below is what I've learned about the parts nobody mentions, mostly from building and rebuilding a noise app used by people who are awake at hours they did not choose.

What brown noise actually is, in one paragraph

Brown noise has energy that falls by about 6 dB per octave as frequency rises — its power spectral density is proportional to 1/f². Pink noise falls at roughly 3 dB per octave; white noise is flat. That single difference is why brown noise sounds like a distant waterfall or a wide, low rumble, and white noise sounds like a hiss or an untuned radio. Most of brown noise's energy sits low, below a few hundred hertz. Which leads directly to the problem almost nobody tells you about.

The low-end tax: why brown noise for sleep sounds different on every speaker

Here is the pattern I'd most like other people to name and design around.

The low-end tax: a small speaker silently discards everything below its roll-off. Because brown noise keeps most of its energy down there, a phone speaker delivers you something closer to pink noise — and you compensate by turning it up, which adds the midrange you were trying to avoid.

A phone's built-in speaker is a few millimetres of moving air. It produces very little output in the lowest octaves, no matter what you feed it. So the brown noise you carefully chose arrives at your ear as a thinner, brighter version of itself. It still masks, but it masks differently: more hiss, less rumble. Then you nudge the volume up, because it isn't doing the job, and now the mid frequencies — the band your ear is most sensitive to, the band a baby's cry and a creaking floorboard live in — are louder than you intended.

Practical consequences:

• On a phone speaker, brown noise is a preference, not a spectrum. If it sounds better to you than white, use it, but don't expect the deep rumble you heard on headphones.
• On any speaker with a real driver — a bedside Bluetooth speaker, a soundbar, laptop speakers with a woofer — brown noise gets much closer to what it's supposed to be, and you'll usually find you need less volume for the same masking.
• Headphones are the honest case, and also the case where the loudness you're choosing is entirely on you.

If you've ever wondered why brown noise feels transformative on one device and pointless on another, this is usually why. It isn't the recording. It's the speaker throwing away the part that makes it brown.

Set the volume against the disturbance, not against comfort

The instinct is to set the volume to a level that sounds nice while you're lying there awake. That's the wrong reference point, because a quiet room at midnight is not the room you'll be woken in.

A better method: think in terms of a masking budget. You are trying to lift the floor to just under the level of the thing that bothers you. So:

1. Identify the actual disturbance. Traffic? Plumbing? A hallway conversation? Your own tinnitus?
2. Play the noise and turn it up until that specific sound stops drawing your attention — not until it's inaudible, just until it stops being an event.
3. Stop there. Then, if anything, come down a notch.

Two things this method makes obvious. First, steady noise masks steady sounds well and transients badly. A hum, a fan, distant road noise, breathing: all easy. A door slamming, a bin lorry, a dog barking once: these have a fast rise time and a level far above the floor, and you cannot mask them without a volume you'd never want in a bedroom. If your problem is bangs, noise is the wrong tool and you're looking for a building fix.

Second, if you find yourself creeping the volume up night after night, that's information. It's easier to add level than to subtract it, and low-frequency sound doesn't nag at you the way a hiss does, so brown noise in particular tends to drift upward unnoticed. Every couple of weeks, start from silence and re-find the level rather than resuming yesterday's.

I'm not going to give you a decibel number for a child's room. That's a safety question, and it belongs to people qualified to answer it, not to me.

Silence is an event: why the fade-out matters more than the sound

The most common way a sleep timer ruins a night is by working exactly as specified. Noise plays for 45 minutes, then stops. The stop is not the removal of a stimulus. The stop is a stimulus. Your auditory system is built to notice change, and an abrupt drop of 20-something dB in the room's floor is a large, fast change.

This is the second pattern worth naming: the silence event. Ending sound abruptly is itself a wake cue, so a timer's job is not to stop the audio but to make the return to the room's real noise floor slow enough that nothing registers as a change.

When we built the timer in Tysha, the fade is long and shaped rather than linear — the last stretch of the ramp is deliberately slow, because the quieter the noise already is, the more a given step down is perceptible as a change. A linear fade to zero has its most noticeable moment at the very end, which is precisely the wrong place to put it. If your current app hard-cuts at the end of a timer, either lengthen the fade or don't use the timer at all and let it run.

The same logic applies at the start. Coming in at full level is a change; coming up over twenty or thirty seconds is not.

Loop seams and the 30-second tell

Most free noise tracks are short recordings on repeat. Even a good loop leaves a seam: a tiny discontinuity, or a slight periodicity in the texture, that repeats at a fixed interval. In daylight you will never hear it. In bed, with nothing else to attend to, some people latch onto it and then cannot un-hear it. The complaint usually arrives as "it sounds like there's a pattern in it", and there is.

That's why Tysha generates its noise in real time rather than shipping audio files. It costs more work — you have to get the filter and the gain staging right, and you have to keep it running reliably while the screen is off — but it removes an entire category of 3am irritation, and it means the fades and the mixing happen to live sound rather than to a clip. If you're picking from what's already on your phone, a generated source is worth preferring; failing that, prefer longer files over short ones.

A setup that survives the whole night

Most bad nights with noise are logistics, not audio. The list that actually matters:

• Nothing that can serve an ad. A pre-roll or mid-roll at 4am is the loudest thing that will happen in your house that night. This is the single strongest argument for playback that doesn't touch the network.
• Offline, so a dropped connection can't stop it. Streaming means one bad Wi-Fi moment ends the masking, which triggers a silence event you didn't schedule.
• Check your alarm is still audible over the level you've chosen. Test this once, while awake.
• Do not disturb on, but with your exceptions set before you rely on it.
• Power. Continuous audio all night drains a phone. Plug it in, and put the cable somewhere you can find it in the dark.
• Position the speaker away from your head, ideally between you and the noise source. You want the room lifted, not your ear.

What brown noise won't do

It won't make a quiet room quieter. It won't shorten the time it takes you to fall asleep if what's keeping you awake is a racing mind rather than sound — although some people find a steady floor gives their attention something uneventful to rest on, which is not the same claim. It won't mask sudden loud transients. And it won't work through a speaker that can't reproduce it, which is the failure mode I see mistaken for "brown noise doesn't work for me" more than any other.

Try it on the best speaker in the room before you decide.