Night review - 17 April
Near-complete sleep coverage from the ceiling mic. Pierre's anchors used for timing.
Recording coverage vs your sleep
- File 6 (22:14 -> 00:40, 2h 26m): early-night sleep. Includes 20 min of pre-sleep wakefulness at start.
- Gap (00:40 -> 00:55, 15 min): recorder off between files. You likely didn't notice File 6 stopping.
- File 5 (00:55 -> 05:34, 4h 39m): late-night sleep. Auto-stopped at 1 GB cap 4 min before wake.
The two sleep files, visualised
| File | Wall-clock | Duration | Events | Events / h | % active |
|---|---|---|---|---|---|
| File 6 (early) | 22:14 -> 00:40 | 2h 26m | 568 | 233 | 52% |
| File 5 (late) | 00:55 -> 05:34 | 4h 39m | 851 | 183 | 65% |
| Total | ~7h 5m audio | 7h 05m | 1,419 | 200 | ~59% |
~59% active across the two sleep files. That means the ceiling mic detected sound above local noise baseline for more than half every second of sleep. For heavy snoring that's plausible; a normal sleeper would be < 20%. Rough translation: a snore or breath burst about every 18-20 seconds, steadily, through the night.
Last night vs your 2mm baseline (Oura)
| Metric | 17 Apr | 2mm avg | Delta | What it means |
|---|---|---|---|---|
| Total sleep (min) | 378 | 436 | -58 | You slept ~1h less |
| Deep sleep (min) | 46 | 66 | -20 | Below the 70-min UARS target |
| REM sleep (min) | 58 | 84 | -26 | Tracks with the shorter night |
| Restless periods | 267 | 324 | -57 | Fewer arousals - good sign |
| HRV (ms) | 47 | 46 | +1 | Flat |
| Lowest HR (bpm) | 51 | 50 | +1 | Flat |
Restless-periods dropping is the most interesting single signal - if it holds over the next 2-3 nights at 2mm, that would be the first concrete Oura-visible sign the MAD is reducing arousal burden (the UARS-specific metric that matters). Everything else this night was just a shorter sleep.
Compared to SnoreLab
| Source | Events / night | What it measures |
|---|---|---|
| SnoreLab phone (baseline to 2mm) | ~20 | Events loud enough to pass SnoreLab's internal intensity gate |
| Ceiling mic, both sleep files combined | 1,419 | Anything above local noise baseline + 1.5σ |
~70x more events on the ceiling mic. SnoreLab only clips loud, distinct snoring bursts - the ceiling mic catches every breath that rises above the room's noise floor. Neither is "right"; they measure different things. SnoreLab is the better intensity signal; ceiling mic is the better coverage signal for eventually correlating against Oura arousals.
Tonight - what to tighten
- Start earlier, stop once. Last night's 15-min mid-night gap happened because File 6 auto-stopped at 1 GB and you didn't restart until 15 min later. Ideal: hit record at 22:00, let it run, deal with the 1 GB splits later.
- Know about the 1 GB cap. At 16 kHz stereo PCM_16 each file holds 4h 39m. Going mono halves the rate (so one file covers 9h 18m - a whole night in one file).
- Set the Vormooi clock. File 6's filename `20260417101433` says 10:14 AM on 17 April, but you claim it stopped at 00:40 on 17 April. That's a ~9.5 hour clock drift. Setting the clock correctly makes filename-derived timestamps reliable, so we don't have to rely on your recollection for every file.
- Mono over stereo. You only have one mic. Stereo doubles file size for no information.
Known loose ends
- File 6's filename doesn't match its stop time. Likely Vormooi clock was set wrong when File 6 was recorded, then corrected before File 5. The audio is fine; just trust your memory over the filename for File 6.
- Fragment 4 (`20260417030611`, 57s) appears to sit at 03:06 inside File 5's recording window. Could be a bookmark / marker file the Vormooi writes without stopping the main recording, or the clock was also off for that fragment. Non-critical - 57 seconds of data.
Generated 2026-04-17. Source data: audio_raw/ceiling-mic/manifest.csv, activity_scan.csv, oura_recent.json, colab_outputs_v3/panthera_uars_v3.parquet. Timings per Pierre's anchors (File 5 stopped 05:34; File 6 started 22:14 night-before). Not medical advice.