Bottom line up front: if your baby sleeps through a barking dog but wakes the moment the sound machine clicks off, the trigger is the change in the room, not the quiet. A running machine holds the noise floor high enough that ordinary house sounds stay buried in it. When it stops — on a timer, at the end of an app routine, or when a phone locks itself — the same creaks and pipes suddenly sit well above a near-silent baseline and become audible for the first time all night. The fix is usually one setting, not a new machine.
I own three sound machines, run two nightly, and have measured all of them where the crib mattress is. Affiliate links never change my verdicts.
The masking math, in one paragraph
This is a signal-to-noise problem, and it is arithmetic rather than mysterious. A machine running at the usual target — roughly 50 A-weighted decibels at your baby’s ear, the figure I work through in how loud white noise should be — sets the floor of the room, and a quiet bedroom with nothing running sits well below that. A stair creak, a toilet flush two rooms away or a car door in the street arrives at the crib at whatever level it arrives at. The only thing that changes is how far it rises above the floor beneath it.
With the machine on, that creak clears the floor by a little: present, unremarkable, ignorable. With the machine off, the identical creak clears a much quieter floor by a much larger margin. Nothing about the creak changed. The contrast did. That is the whole phenomenon, and it explains what parents find maddening — the vacuum at 2 p.m. is fine, the fridge compressor at 2 a.m. is not.
A second effect stacks on top. Babies surface into light sleep repeatedly overnight, and the AAP’s HealthyChildren.org notes both that a six-month-old normally wakes and resettles within a few minutes, and that babies who fall asleep under particular conditions “may struggle to go back to sleep if they wake up during the night.” A machine that switched off at 11 p.m. means the 1 a.m. surfacing happens in a room that no longer resembles the one they fell asleep in.
The three ways your sound stops without you deciding
Almost every case I have helped a friend debug came down to one of these, and two are defaults nobody chose.
An auto-shutoff timer you didn’t know was on. Plenty of machines offer one, and some remember it once you have used it. Sixty minutes is a common interval, which lands the shutoff squarely at the end of the first deep stretch — right before the first real surfacing.
A routine duration in an app. Connected machines have a schedule rather than a physical timer, and the sound plays for as long as the routine you built says it does — a routine tapped early also ends early. It is a configuration issue rather than a fault, and our Hatch Rest review covers what that app does and does not control.
A phone doing phone things. A streamed track ends, an autoplay queue moves to something with a melody, a notification ducks the audio, or the screen locks and the app suspends. Phones are the least reliable sound source in the house.
What the machines actually publish
Very few manufacturers tell you what happens to the sound over eight hours. Two of them do.
The Yogasleep Dohm Classic is mechanical: an actual fan spinning inside a shell, with two speeds and a tone collar. Yogasleep’s Dohm Classic page calls the sound “fan-based, natural, and non-looping,” made by a real fan motor “without any digital audio,” and lists it at $49.99. It has no timer and no schedule — on or off — so the failure mode this article is about cannot happen to it.
The LectroFan Classic is the digital answer: twenty sounds, described on the manufacturer’s listing as “ten non-repeating unique fan sounds, and ten non-repeating unique white noise sounds,” synthesized rather than played from a file. It does carry a sleep timer that shuts the unit down “after you’ve drifted off to sleep” — exactly the feature to leave switched off if your baby is a silence-waker.
Everything else is a guess. Most machines publish neither their output level nor whether the audio is a looped file, which is why I treat “non-looping” as a spec worth paying for. My best white noise machines roundup ranks on those two things rather than on features the baby cannot use.
The five-minute audit
- Open your machine’s manual or app and find the words timer, duration or auto-off. Turn it off, or set the sound to play until you stop it.
- Sit in the room at night with the machine off and listen for two minutes. Whatever you can hear is what your baby hears the instant the sound stops.
- If that is a boiler, a road or a sibling’s door, the machine should run all night rather than to a timer.
- Re-check the volume once it is continuous. All night at high volume is the combination to avoid; all night at a measured, across-the-room level is not.
- Write the setting down. Dials only ever travel one direction.
Step four is an honest tension in my own advice. Duration and level interact, so a timer genuinely is the more conservative default if you are never going to measure. If you want eleven hours of sound, the price of admission is a meter reading rather than a guess.
Then stop tiptoeing
The other half of this, and the part that saves whole naps: once the sound floor is steady, the household above it does not need to whisper. A house that goes abnormally silent for naps turns every unavoidable noise — the doorbell, the older child — into an event.
Load the dishwasher. Talk at a normal volume. The machine is doing the work.
When the machine isn’t the culprit
Not every waking is acoustic, and swapping hardware to fix a scheduling problem is an expensive way to learn that. A 5 a.m. start is more often a daytime-sleep or bedtime-timing problem — Betteroo’s guide to a baby who wakes too early walks through the usual causes in order, and recommends steady white noise specifically to mask dawn birds and traffic. If wakings are spread through the night rather than clustered near dawn, timing is the likelier suspect.
Sorting an environment problem from a schedule problem is what the free two-minute sleep triage Betteroo runs is for, and it costs nothing to run before you buy anything.
FAQ
Should white noise run all night or turn off after my baby falls asleep?
If your baby wakes when the sound stops, run it all night. The trade-off is exposure: duration and volume matter together, so all-night sound is only reasonable at a low, measured level with the machine across the room. If you would rather not measure, a timer is the more cautious setting and you accept the wakings that come with it.
Will running white noise all night make my baby dependent on it?
Not in any way that costs you anything. It runs without your involvement, nobody has to reapply it at 3 a.m., and plenty of adults use it for life. I worked through the argument in weaning off white noise: volume deserves auditing far more than the habit does.
Why does my baby sleep through loud noises but wake in a quiet room?
Because the ear responds to contrast, not absolute volume. A loud sound over a high noise floor stands out less than a soft sound over a near-silent one — the same reason a whisper carries in a library and not at a party.
Does a looping sound file wake babies at the seam?
It can, if the loop point is audible — a hard cut or a fade at the join is a change in the sound environment, the same mechanism as a timer but smaller and repeating. That is why I would rather buy a mechanical or synthesized unit that says “non-looping” in its own specifications than one that stays quiet about how its audio is made.
Is a phone app good enough as a sound machine?
For a nap in a hotel room, yes. As the nightly source, no — screen locks, notifications and autoplay queues all end the sound without warning.
