00-002 — Gentle Awakening Mechanisms

Status and scope

Status: research baseline and product design input, not a medical claim.

Last literature search: 2026-08-20. Product scope revised: 2026-09-06. Camera, bed-presence feedback and automatic bed-exit dismissal are removed. The calibrated envelope below remains a design target, not a claim of deployed behavior. The current two-stage timing contract is recorded in the wake-session specification; the scope change does not itself retime that audio implementation.

This specification defines what Robine Wake means by a gentle awakening, which mechanisms are supported by human research, and how those mechanisms should be translated into testable product behaviour. It does not prescribe one allegedly “scientific” volume curve: the literature does not establish one.

The extracted scientific dataset — study design, sample, intervention, outcomes, statistics, and limitations — is maintained separately in evidence-review.md. That dataset is the evidentiary source; the engineering rules in this file are explicitly downstream interpretations.

The review prioritised peer-reviewed human studies, systematic or structured reviews, and studies comparing an intervention with a control condition. Searches covered sleep inertia, artificial dawn, alarm acoustics, music and melody, alarm loudness, sleep-stage alarms, snoozing, temperature, vibration, multimodal alarms, and sensory awakening. PubMed and reference lists of the principal reviews were used. This is an evidence map for product engineering, not a registered systematic review, and cannot guarantee that every publication in every language was found.

Product definition

“Gentle” has three independent outcomes. A design can improve one and worsen another, so they must never be collapsed into a single score:

1. Wake reliability — the sleeper becomes behaviourally awake before the hard deadline.
2. Wake experience — the transition is perceived as comfortable and not startling or aversive.
3. Post-wake function — subjective sleepiness and objective vigilance recover after awakening.

The alarm must also preserve sleep before the requested wake time. An intervention that creates repeated micro-awakenings may appear to reduce sleep inertia by starting the awakening early while actually reducing sleep efficiency.

Physiological model

Sleep inertia is a temporary state of hypoarousal after awakening. It can include sleepiness, confusion, negative affect, reduced vigilance, slower responses, and a desire to return to sleep. Brain recovery is heterogeneous rather than an immediate global switch: sleep-like and wake-like activity can coexist during the transition.

Its severity is affected by at least:

• circadian phase — controlled laboratory work found cognitive sleep inertia substantially worse during the biological night than the biological day;
• homeostatic sleep pressure — sleep loss and recovery sleep can intensify it;
• time since awakening — different subjective and objective functions recover at different rates, sometimes over tens of minutes or longer;
• sleep stage — some studies report worse outcomes after NREM or slow-wave sleep while others find no independent stage effect;
• person and context — chronotype, sleep debt, age, health, habitual behaviour, and task all contribute large variability.

Robine Wake does not measure sleep stage, circadian phase, alertness or fitness to perform a safety-critical task. An explicit button response is not proof that sleep inertia has dissipated.

Evidence map

Confidence describes the evidence for a consumer bedroom alarm, not the quality of the researchers or the plausibility of the mechanism.

Mechanism	Main finding	Confidence	Product consequence
sufficient sleep and regular timing	sleep pressure and circadian phase strongly modulate inertia	moderate	do not promise that alarm shaping can compensate for sleep deprivation
30-minute artificial dawn	several small trials improve subjective alertness; objective effects are mixed	low–moderate	best-supported future environmental feature, but optional
bright or blue-enriched light only after waking	results are mixed and task-dependent	low	maintain useful room light after waking; do not market blue content as a proven cure
melodic alarm sound	small field studies show fewer vigilance lapses and false starts	low	prefer clearly melodic, user-selected material in the gentle phase
rhythm or tempo alone	a 105 BPM rhythmic treatment did not outperform control	low	do not infer efficacy from BPM alone
preferred music	limited studies suggest benefits to comfort, sleepiness, or response speed	low	expose sound choice and learn preference; avoid one universal “best” track
gradual audio loudness	no controlled evidence establishes an optimal duration or curve	insufficient	treat the envelope as a product hypothesis and calibrate/test it
low-frequency tone or voice	wakes children and emergency-study populations more reliably than high-pitched alarms	moderate for emergency arousal; low for this product	consider for the strong fallback, not as proof of a pleasant adult awakening
waking in “light sleep”	stage effects are inconsistent and practical stage estimation is imperfect	insufficient	never delay the hard deadline; do not claim optimal-stage waking
snoozing	two small laboratory studies conflict; habitual snoozers may differ	low	make snooze explicit and measurable, not the default gentle mechanism
temperature manipulation	promising subjective effects in older small studies; multimodal trial was null overall	insufficient	defer until independently testable hardware exists
vibration	can wake sleepers and may be pleasant, but evidence targets prototypes or emergency/clinical use	low	optional secondary channel, not a replacement for the reliable audio fallback
odour	human olfaction does not reliably wake sleepers	moderate	exclude odour as a safety or wake-reliability mechanism
multimodal smart bedroom alarm	recent within-subject trial found no overall vigilance, sleepiness, or mood benefit and slightly lower sleep efficiency	low	add modalities independently and measure sleep disruption as well as benefit

Interpretation by mechanism

Light before waking

Artificial-dawn studies commonly start light about 30 minutes before scheduled waking and rise toward roughly 100–300 lux at the eye. Trials in people who report difficulty waking have repeatedly found better subjective awakening, mood, or alertness. Some very small studies also report better reaction time or cognitive performance. Other trials find subjective benefits without performance or cortisol effects, and a study in older adults found no alertness benefit while detecting more wake before the alarm.

The likely effect is acute and may involve increased pre-wake arousal and thermoregulation. A shift in melatonin timing and the cortisol awakening response are not established explanations. Recent work also questions whether awakening itself causes the previously assumed burst in cortisol secretion.

Design rule: if light hardware is added, the default product profile is a 30-minute dawn. Light duration must not be shortened because a non-EEG sensor labels the user as being in light sleep. Record intervention start and explicit pre-alarm user interactions, without inferring sleep or wake state, and allow the intervention to be disabled.

Sound identity

The strongest product-relevant auditory signal is preliminary evidence for melody, not loudness shape. A small ecological experiment found fewer attention lapses and false starts after a melodic treatment, but no significant mean reaction-time benefit; rhythm alone did not help. The broader auditory review found too little adult evidence for firm prescriptions, although melodic alarms and preferred music are promising.

Emergency studies answer a different question: whether a sleeper wakes at all. They support low-frequency signals and voice over conventional high-frequency beeps in children and several special populations. They do not justify playing a 100 dB signal in a bedroom or claiming that the same signal reduces sleep inertia.

Design rule: the gentle phase should use a user-selected melodic sound with enough spectral energy to remain audible on the installed speakers. The strong phase optimises wake reliability and distinctness. Sound metadata such as “melodic”, “salient”, and measured loudness is more useful than an unvalidated “gentleness” label alone.

Loudness and the missing curve evidence

Published alarm work generally compares sound type or increases test intensity to measure an awakening threshold. The recent multimodal experiment used a two-minute increase to 50% of an unspecified device maximum, but the combined intervention did not improve its primary outcomes overall. This does not validate a two-minute curve. No identified controlled study compared linear, exponential, logarithmic, sigmoidal, or stepped bedroom-alarm envelopes, nor did one validate Robine Wake’s current five-minute linear logical-volume ramp.

ALSA mixer dB is an electrical gain setting, not sound-pressure level at the pillow. Track mastering, amplifier gain, speaker sensitivity, distance, room acoustics, and background noise all change the audible result. A curve expressed only as mixer values is therefore neither transferable nor perceptually stable.

Design rule: replace curve-first thinking with an audible-envelope model:

• calibrate each bundled track to a common programme loudness;
• measure A-weighted level at the pillow for the actual Amp2 and speaker setup;
• store a personalised audible onset and a separately bounded maximum;
• specify changes in dB, not arbitrary 0–100 percentage points;
• use monotonic, click-free transitions and never exceed the hardware safety cap;
• escalate on elapsed non-response and observed wake behaviour, not because a mathematically smooth curve happens to end.

Robine Wake hypotheses

The published data constrain the design but do not determine it. Robine Wake therefore adopts the following original hypotheses as product decisions. They are not presented as findings from the cited papers.

H1 — Gentleness is controlled arousal, not continuous quietness

An alarm that remains below the sleeper’s audible threshold for several minutes is not gentle; it is absent, followed by a late surprise. Startle is more likely to be related to the level and rate of change when the signal becomes perceptible than to the total mathematical duration of a mixer ramp.

Robine Wake shall cross the likely audibility threshold deliberately in the first seconds, with a bounded slope, then give the brain time to orient before adding more level.

H2 — Stable melody supports orientation

The auditory system needs a stable object to recognise. Continually changing gain for five minutes makes both the sound and its meaning unstable. A melodic plateau after fade-in should provide a less confusing transition than an uninterrupted ramp while preserving the preliminary vigilance benefit associated with melody.

H3 — Salience steps are better than imperceptible micro-increments

One-second mixer changes are operationally smooth but most are not meaningful to a sleeper. Discrete 3 dB steps, separated by stable periods, should renew attention without the abruptness of switching immediately to the strong alarm. The 3 dB value is a psychoacoustic engineering choice, not an outcome from an alarm trial.

H4 — Failure is behavioural, not volumetric

For this camera-free product, absence of response means no authorized explicit snooze or dismissal before the deadline. Elapsed non-response triggers the strong fallback. No movement, occupancy or inferred sleep state can pause escalation; reaching the tactile ceiling cannot change the audio deadline.

H5 — The fallback changes meaning before maximising level

A novel, salient sound should regain attention more efficiently than making the same familiar track progressively louder. The strong phase therefore changes sound identity and spectral character, then uses a higher but bounded level. It is not the last point of the gentle curve.

H6 — The bedroom, not the mixer, defines loudness

Raw ALSA dB does not describe the pressure reaching the pillow. Background noise, track mastering, speaker sensitivity, placement, distance, and room reflections all matter. Robine Wake’s target is an acoustic level at the pillow; the Amp2 mixer is only the actuator used to reach it.

H7 — Light prepares; sound declares

Future visible-light hardware should begin a slow environmental transition before the requested time. Audio remains the unambiguous declaration that wake time has arrived. Visible-light and tactile preparation remain independently configurable.

Calibrated Robine Wake envelope — design target

This is the proposed calibrated profile derived from the hypotheses above. Its implementation and acoustic acceptance remain separate from the camera-removal and tactile work; current timing remains defined by 20-002 until that profile is implemented. The table below must not be combined with the current pre-wake ramp and five-minute plateau as though they were one timeline.

Let:

• B be the A-weighted background level measured near the pillow;
• L_gentle = clamp(B + 10 dB, 42 dBA, 52 dBA);
• L_strong = clamp(max(L_gentle + 12 dB, 62 dBA), 62 dBA, 72 dBA);
• S(u) = 3u² - 2u³, the smoothstep function for 0 ≤ u ≤ 1;
• G(L) be the calibrated Amp2 mixer gain required to produce acoustic level L.

Until physical SPL calibration exists, the implementation may approximate the relative changes in mixer dB, but the UI must not label those values as dBA.

Time from scheduled wake	State	Output
0–12 s	gentle_fade	gentle track rises from L_gentle − 18 dB to L_gentle using S(t/12) in the dB domain
12–72 s	gentle_hold	hold the melodic track at L_gentle
72 s	gentle_step_1	increase to min(L_gentle + 3 dB, L_strong − 3 dB)
102 s	gentle_step_2	increase to min(L_gentle + 6 dB, L_strong − 3 dB)
132 s	gentle_step_3	increase to min(L_gentle + 9 dB, L_strong − 3 dB)
162–166 s	strong_crossfade	crossfade over 4 s to strong_sound_id and L_strong
after 166 s	strong_hold	loop the strong sound at L_strong until dismissal, snooze, or safety timeout

The fade is S-shaped in acoustic dB, not in the 0–100 UI value and not in raw PCM amplitude. This gives zero slope at both ends, avoids a click or sudden change of derivative, and reaches a useful level quickly. The later progression is stepped because the product is trying to renew salience, not draw an aesthetically smooth graph.

Explicit response only

Only an authorized snooze or dismissal interrupts the escalation timeline. No movement-based extension exists. In the proposed calibrated profile, the strong crossfade therefore starts at 162 seconds with no former 60-second extension. For the current implementation, retain the deadline in 20-002. Tactile output follows the same session clock and never postpones either deadline.

Track preparation

Bundled gentle tracks should be normalised to approximately −16 LUFS-I with a true-peak ceiling of −1 dBTP. Strong tracks should use the same loudness reference so that the acoustic transition is controlled by the envelope rather than arbitrary mastering differences. These are production hypotheses chosen for consistency, not clinical thresholds.

Timing and sleep-stage alarms

Sleep stage is scored clinically from signals such as EEG, eye movement, and muscle tone. Consumer and contactless devices are generally much better at recognising sleep than wake and show inconsistent stage classification. One published smart alarm predictor reported only 66–70% four-class accuracy and did not establish a clinically meaningful improvement in morning function.

Robine Wake has no camera or presence sensor and must not label N1, N2, N3 or REM. Any prelude is determined by explicit user settings and time, never an inferred sleep stage or detected movement.

Snooze

A 2022 pilot with ten participants suggested snoozing extended light sleep near waking and could prolong inertia. A later within-subject study of 31 habitual snoozers found that 30 minutes of snoozing improved or did not affect immediate cognitive performance, lost about six minutes of sleep, and produced no clear sleepiness, mood, cortisol, or architecture effect. The samples and populations do not support one universal policy.

Design rule: retain an explicit snooze limit, record the number and total sleep fragmentation, and let habitual snoozers compare snooze-enabled and direct-wake profiles. Never describe snoozing as proven beneficial or harmful for everyone.

Confirmation versus alertness

Record explicit snooze and authorized dismissal as user interactions. The system does not confirm bed exit or cognitive alertness. A physical response acknowledges the alarm, without proving that sleep inertia has ended.

Evidence-informed wake sequence

optional light prelude
        |
scheduled audio + optional progressive vibration
        |
        +-- authorized physical dismissal --> stop outputs
        +-- authorized snooze --> silence outputs --> restart
        +-- no explicit response --> strong audio at its fixed deadline
                                      |
                                      +--> dismissal / snooze / safety timeout

The vibration mode is defined in bed vibration. Its failure does not suppress or delay the audible fallback. No presence signal participates in this sequence.

Product diagnostics

Each wake session should eventually support the following privacy-preserving event data, without collecting images or presence observations:

• scheduled time, actual audio onset, and audio-start latency;
• envelope segment, sound ID, normalised track loudness, mixer dB, and calibration profile identifier for every transition;
• explicit interaction, snooze, escalation, dismissal and timeout;
• tactile mode, calibrated intensity, limiter/cutoff events and failures;
• minutes of any pre-wake intervention and explicit pre-alarm interactions;
• the configured acoustic calibration and the fallback deadline applied.

Raw camera frames, inferred medical conditions, and definitive sleep-stage labels remain out of scope.

Commissioning constraints

Commissioning is a hardware verification step, not a human-subject experiment. It must confirm that both installed speakers reproduce the requested levels at the pillow position, that transitions are click-free, that the limiter prevents output above 72 dBA under the calibration track, and that the strong fallback still starts on time with the tactile output disconnected or failed.

Requirements

• GWA-001 — Product language shall distinguish wake reliability, wake experience, and post-wake function.

• GWA-002 — No audio curve shall be described as evidence-based without a direct controlled comparison supporting that curve.

• GWA-003 — Audio envelopes shall be defined in calibrated dB transitions rather than raw mixer percentages alone.

• GWA-004 — The strong fallback deadline shall be independent of sleep-stage inference and tactile availability.

• GWA-005 — Retired: interpretation of infrared observations.

• GWA-006 — Retired: interpretation of detected bed exit.

• GWA-007 — Planned calibrated audio profile: implement the fade–hold–step–fallback envelope after its separate implementation and acoustic validation; current runtime timing is specified by 20-002.

• GWA-008 — Any light prelude shall be user-controlled and shall not move the hard wake deadline.

• GWA-009 — Retired: movement-based escalation extension is removed.

• GWA-010 — The interface shall label research-derived facts separately from hypothesis-driven product settings.

• GWA-011 — Only authorized explicit interactions shall interrupt escalation; no presence or movement signal shall delay the selected strong deadline.

• GWA-012 — User acknowledgement shall not be presented as proof of bed exit or recovered cognitive alertness.

Research references

Sleep inertia and its modifiers

1. Ruby PM et al. From physiological awakening to pathological sleep inertia. Neurophysiol Clin. 2024. doi:10.1016/j.neucli.2023.102934
2. Jewett ME et al. Time course of sleep inertia dissipation in human performance and alertness. J Sleep Res. 1999. PubMed 10188130
3. Scheer FAJL et al. An endogenous circadian rhythm in sleep inertia results in greatest cognitive impairment upon awakening during the biological night. J Biol Rhythms. 2008. doi:10.1177/0748730408318081
4. Silva EJ, Duffy JF. Sleep inertia varies with circadian phase and sleep stage in older adults. Behav Neurosci. 2008. doi:10.1037/0735-7044.122.4.928
5. Gorgoni M et al. Sleep inertia and sleep homeostatic processes: measurement of electroencephalographic awakening reaction and task performance. J Sleep Res. 2015. PubMed 26004680
6. Hilditch CJ, Dorrian J, Banks S. Time to wake up: reactive countermeasures to sleep inertia. Ind Health. 2016. doi:10.2486/indhealth.2015-0236
7. Chen S et al. Thalamic dynamics orchestrate the recovery of tonic alertness during nocturnal sleep inertia. Commun Biol. 2026. PubMed 41832350

Light and artificial dawn

8. Thorn L et al. The effect of dawn simulation on the cortisol response to awakening in healthy participants. Psychoneuroendocrinology. 2004. doi:10.1016/j.psyneuen.2003.08.005
9. Giménez MC et al. Effects of artificial dawn on subjective ratings of sleep inertia and dim light melatonin onset. Chronobiol Int. 2010. doi:10.3109/07420528.2010.496912
10. van de Werken M et al. Effects of artificial dawn on sleep inertia, skin temperature, and the awakening cortisol response. J Sleep Res. 2010. doi:10.1111/j.1365-2869.2010.00828.x
11. Thompson A et al. Improving the health of office workers: effects of artificial dawn on morning physiology and performance. Eur J Appl Physiol. 2014. doi:10.1007/s00421-014-2831-z
12. Santhi N et al. Morning sleep inertia in alertness and performance: effect of cognitive domain and white light conditions. PLoS One. 2013. PubMed 24260280
13. Hilditch CJ et al. Rise and shine: the use of polychromatic short-wavelength- enriched light to mitigate sleep inertia at night following awakening from slow-wave sleep. J Sleep Res. 2022. doi:10.1111/jsr.13558
14. Gabel V et al. Effect of artificial dawn light on cardiovascular function, alertness, and balance in middle-aged and older adults. Sleep. 2020. doi:10.1093/sleep/zsaa082
15. Norden MJ, Avery DH. A controlled study of dawn simulation in subsyndromal winter depression. Acta Psychiatr Scand. 1993. doi:10.1111/j.1600-0447.1993.tb03415.x
16. Zhu YY et al. Effects of artificial dawn on sleep inertia in alertness, mood, and cognitive performance in healthy participants. Sleep. 2018 conference abstract. doi:10.1093/sleep/zsy061.209
17. Bowles NP et al. Awakening not associated with an increased rate of cortisol secretion. Proc R Soc B. 2025. PMC 11732391

Sound, alarm signals, and vibration

18. McFarlane SJ et al. Alarm tones, voice warnings, and musical treatments: a systematic review of auditory countermeasures for sleep inertia in abrupt and casual awakenings. Clocks Sleep. 2020. doi:10.3390/clockssleep2040031
19. McFarlane SJ et al. Auditory countermeasures for sleep inertia: exploring the effect of melody and rhythm in an ecological context. Clocks Sleep. 2020. doi:10.3390/clockssleep2020017
20. McFarlane SJ et al. Alarm tones, music and their elements: analysis of reported waking sounds to counteract sleep inertia. PLoS One. 2020. PubMed 31990906
21. Bruck D et al. How does the pitch and pattern of a signal affect auditory arousal thresholds? J Sleep Res. 2009. doi:10.1111/j.1365-2869.2008.00710.x
22. Smith GA et al. Comparison of the effectiveness of female voice, male voice, and hybrid voice-tone smoke alarms for sleeping children. Acad Pediatr. 2019. doi:10.1016/j.acap.2019.06.016
23. Bruck D, Thomas IR. Smoke alarms for sleeping adults who are hard-of-hearing: comparison of auditory, visual, and tactile signals. Ear Hear. 2009. doi:10.1097/AUD.0b013e3181906f89
24. Korres G et al. A vibrotactile alarm system for pleasant awakening. IEEE Trans Haptics. 2018. doi:10.1109/TOH.2018.2804952
25. Attali V et al. Awakening efficacy of a vibrotactile device in patients on home nocturnal ventilatory assistance and healthy subjects as family caregiver proxies. Chron Respir Dis. 2021. doi:10.1177/1479973120983331

Smart alarms, snooze, and multimodal interventions

26. Slyusarenko K, Fedorin I. Smart alarm based on sleep stages prediction. EMBC. 2020. doi:10.1109/EMBC44109.2020.9176320
27. Campanella A et al. The efficacy of a multimodal bedroom-based “smart” alarm system on mitigating the effects of sleep inertia. Clocks Sleep. 2024. doi:10.3390/clockssleep6010013
28. Tsutsui K et al. Effects of using a snooze alarm on sleep inertia after morning awakening. J Physiol Anthropol. 2022. PMC 9804954
29. Sundelin T et al. Is snoozing losing? Why intermittent morning alarms are used and how they affect sleep, cognition, cortisol, and mood. J Sleep Res. 2024. doi:10.1111/jsr.14054
30. Lee YJ et al. Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis. J Clin Sleep Med. 2025. doi:10.5664/jcsm.11460
31. Chinoy ED et al. Performance of seven consumer sleep-tracking devices compared with polysomnography. Sleep. 2021. PubMed 33378539
32. Carskadon MA, Herz RS. Minimal olfactory perception during sleep: why odor alarms will not work for humans. Sleep. 2004. PubMed 15164891