Deep Sleep is the heavier sibling of Sleep mode, built for the nights an ordinary preset still feels like it's happening a little too close to the surface. Every sound below is an Official Sound DNA recipe, composed live rather than played back from a file, and all of them share the same core traits: a lower, wider drone, a longer reverb tail, bells rare enough to almost not register at all. What actually varies between them is mostly texture (rock, water or rain) and how far into the room each one reaches.
A handful of these lean on stone and enclosed space rather than water. Cavern and Hollow Earth both use a cave texture with an occasional buried drip, but Hollow Earth pushes it further, running the longest reverb decay on the site for the sense of the biggest enclosed space available. Fathoms sits between the two, pairing that same cave texture with brown noise instead of waves. On the water side, Ocean Deep, Beneath the Waves and Slow Tide all vary the same idea: Ocean Deep favors more wave and less of everything else, for anyone falling asleep to the idea of being near water rather than under it; Beneath the Waves goes bigger and slower, with a wider stereo image and less brown noise underneath; Slow Tide pulls the swell back further still, almost nothing left at the high end besides the drone and an occasional wave. Rain on Stone and Deep Rain both go heavier on rain specifically. Rain on Stone runs denser and wetter in the reverb, reading like rain landing on something harder than a roof; Deep Rain runs slower, with drops landing further apart under a noticeably thicker drone. And for the lowest setting on the site, Abyss and Undertow both go there: Abyss keeps the drone low enough to stop registering consciously, Undertow pairs that same heavy drone with a faster wave swell and a cave texture in place of the usual brown-noise base.
Deep Sleep earns the name mostly in the numbers. Regular Sleep mode runs five detuned drone voices under a nine-second reverb tail; Deep Sleep pushes that to seven voices and twelve seconds, with a filter that never opens past 460Hz, so almost nothing in the mix sits above the range a room's own hum already occupies. Stereo width is also set to its widest anywhere on the site here, wider than Sleep, Relax or Focus. Combined with that long reverb tail, a low-frequency source stops reading as coming from one direction and starts reading as filling the room instead, closer to how an actual building's ambient hum behaves than a single point-source sound ever does.
There's no wrong choice here in the sense that matters. Every preset stores its score rather than an audio file, so a Deep Sleep session generates new material for as long as it plays, with no seam to wake up during and nothing to load past the first few seconds. If it turns out to be heavier than what you actually need most nights, regular Sleep mode is one step lighter on the same engine, with more bell activity and a shorter reverb tail.