No watch. Just a phone on the nightstand.
Phone charging all night, app open — and for naps, the phone right there with you. A different sensor kit than the wrist, with one surprise: the phone can hear you breathe.
The sensors, ranked honestly
| Channel | What it gives us | The catch |
|---|---|---|
| Sonar (speaker + mic) | The phone plays inaudible chirps and reads your chest movement off the echo — breathing rate to within ~0.1 breaths/min, up to ~1 m away, even under a blanket. The primary night channel; already shipping on iPhone (SleepScore). | Single sleeper only — it measures the bed, not a person. Needs a per-device self-test (a few people can hear the chirp; dogs can). |
| Microphone (passive) | Snore detection — the best-validating pure-phone feature (97–99% specificity in quiet rooms). | Fans and white-noise machines mask it; disabled in noisy bedrooms rather than half-working. |
| Accelerometer | At night: phone-pickup = the one perfect "awake" signal. In a nap, phone on your chest: reads individual breaths at clinical grade (±0.5 s per breath) — a signal the Watch can't give us live at all. | Phone on the mattress sensing scored worst of every modality ever tested (κ 0.07) — mattress mode is cut. |
| Excluded | Camera (would have to aim at your face all night), ambient light (Apple restricts it), UWB/Wi-Fi sensing (no solo-phone API). Ruled out, not forgotten. | |
The nightstand night, drawn out
What phone-only honestly delivers
- A bulletproof deadline alarm — the non-negotiable core, and the phone does it superbly.
- Restless-moment early wake (optional). In a narrow window before your deadline, fire early when you're detectably stirring. Honest label: the one trial of a bedroom smart alarm found no measurable benefit over a good gentle alarm — so this ships narrow, off-by-default-ish, and measured in-product.
- Snore events — in quiet bedrooms.
- A sleep/wake summary with the bias printed on it. Sleep detection is excellent; quiet 3 am wakefulness gets scored as sleep (every consumer device shares this flaw — we say it, they don't).
- Breathing-rate trend — single sleeper only.
What it never claims
- Sleep stages. Best phone staging evidence comes from vendors grading their own homework; phone-accelerometer staging predicted 59% deep sleep where reality was 11%. No hypnogram, ever.
- A validated onset time. No shipping phone app has one. Ours would say "asleep no earlier than…" until our own validation earns more.
- Per-person tracking with a partner or pet in range. One tap declares them; breathing features switch off rather than guessing whose breath it is. You keep the excellent alarm and snore events.
- Silent wake. Audible only — the wrist owns silent. This is the watch's biggest structural win.
The nap case: the phone's quiet superpower
For naps the phone doesn't need the nightstand — on your chest, its accelerometer reads every breath at clinical grade. That's a live respiration waveform the Watch APIs don't offer at any price. The nap itself stays fixed-clock ~20 minutes (a call can suspend sensing mid-nap, and a still-but-awake napper fools every sensor — so the alarm is pre-scheduled and sensing may only ever fire it earlier). Couch-cushion and in-pocket placements can't support any onset claim, so they get the fixed clock, honestly.
The tier order, whole product: Watch → phone-on-nightstand → plain fixed clock. Every tier keeps the identical guaranteed deadline; each higher tier adds refinement, never dependence. Nothing requires buying hardware — the free floor is a genuinely excellent alarm.
Privacy, since the mic runs all night: everything is processed on the device, no audio is stored or uploaded, and the persistent orange mic indicator gets explained on night one. Decline the mic and you get a clean alarm-only mode — not a degraded tracker.