Add curated AquaSynth dings MCP service

docs/research/auditory-notifications/README.md

@@ -0,0 +1,74 @@
+# Auditory Notification Research Distillation
+
+This ledger grounds AquaSynth Dings in auditory-interface research without
+pretending that a desktop agent notification is a medical or aviation safety
+alarm. The local PDFs are research inputs, not redistributed product assets.
+
+## Findings that constrain the implementation
+
+- **Use decaying, percussive envelopes.** Foley et al. found temporally varying,
+  percussive tones less annoying and more detectable than flat tones under
+  concurrent speech load. Every curated instrument therefore has a short attack,
+  brief sustain, and decaying tail. Flat beeps and broadband noise are rejected.
+- **Map urgency through timing deliberately.** Haas and Casali compared 0, 150,
+  and 300 ms inter-pulse intervals; shorter intervals increased perceived
+  urgency without improving detection time. Routine AquaSynth motifs therefore
+  keep onset gaps in the 180–320 ms range. We do not make ordinary progress
+  notifications sound urgent merely to make them noticeable.
+- **Keep pitch and spectral agitation restrained.** Hellier, Edworthy, and
+  Dennis found speed, fundamental frequency, repetition, and inharmonicity all
+  affect perceived urgency. The catalog stays in a calm mid register, uses
+  consonant C-major intervals, limits brightness, and forbids noisy timbres.
+- **Do not rely on arbitrary melodies alone.** Sanderson, Wee, and Lacherez
+  found poor learning and persistent confusion in a standardized melodic alarm
+  set. Here the motif grammar is small and systematic: landing pitch and motion
+  carry event class, while timbre independently identifies the speaking agent.
+- **Expect masking when sounds overlap.** Bolton et al. experimentally showed
+  that masking a tonal alarm's primary harmonic can make it indistinguishable.
+  The service mixes concurrent sounds, but the curated catalog intentionally
+  separates timbre families and exposes `stop_dings`; a later field test should
+  measure identification under realistic overlap before expanding the catalog.
+- **Prefer a small, tested vocabulary.** Studies of IEC melodic alarms and
+  overlapping alarms show that nominal distinctness is not reliable human
+  discriminability. New instruments require listening/identification trials;
+  parameter validation is only admission hygiene.
+
+## Implemented policy
+
+- Event selects motif; instrument selects agent identity.
+- Instrument source is catalog-only. MCP callers cannot inject patch scripts.
+- Root frequency: 196–392 Hz.
+- Brightness curation score: 0.20–0.78.
+- Decay: 0.25–1.50 seconds; attacks remain short and nonzero where practical.
+- No broadband-noise instruments.
+- Multi-note onset gaps: 180–320 ms.
+- One persistent playback mixer owns the output device.
+- Aqua daemon remains the sole synthesis authority.
+
+## Sources
+
+1. Foley, L. et al. (2022), *More detectable, less annoying: Temporal
+   variation in amplitude envelope and spectral content improves auditory
+   interface efficacy*, JASA 151(5), DOI `10.1121/10.0010447`. Saved as
+   `foley-et-al-2022-more-detectable-less-annoying.pdf`.
+2. Hellier, E. J., Edworthy, J., & Dennis, I. (1993), *Improving auditory
+   warning design: quantifying and predicting the effects of different warning
+   parameters on perceived urgency*, Human Factors 35(4), DOI
+   `10.1177/001872089303500408`.
+3. Haas, E. C., & Casali, J. G. (1993), *The Perceived Urgency and Detection
+   Time of Multi-Tone and Frequency-Modulated Warning Signals*, HFES Annual
+   Meeting, DOI `10.1177/154193129303700906`.
+4. Sanderson, P. M., Wee, A., & Lacherez, P. (2006), *Learnability and
+   discriminability of melodic medical equipment alarms*, Anaesthesia 61(2),
+   DOI `10.1111/j.1365-2044.2005.04502.x`.
+5. Lacherez, P., Seah, E. L., & Sanderson, P. M. (2007), *Overlapping Melodic
+   Alarms Are Almost Indiscriminable*, Human Factors 49(4), DOI
+   `10.1518/001872007X215719`.
+6. Bolton, M. L. et al. (2020), *An Experimental Validation of Masking in IEC
+   60601-1-8:2006-Compliant Alarm Sounds*, Human Factors 62(6), DOI
+   `10.1177/0018720819862911`.
+7. NIST Technical Note 1950, *Auditory Warning Signals*. Saved as
+   `nist-tn-1950-auditory-warning-signals.pdf`.
+
+The source README for the event vocabulary is
+<https://github.com/iain/minimal-dings/blob/main/minimal-dings/README.md>.