You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Ternary Hysteresis Memory (configure(hysteresis_q8)) — trits now hold state across the threshold edge; deep sign flips still captured instantly. Verified: -34.3% trit chatter, 90.0% of waves preserved.
Live Threshold Self-Reinforcement (configure(live_sigma)) — thresholds re-derived from running variance every feed; a 16x noise-floor jump: 1530 false beats (fixed) vs 4 (live).
QState.wave_energy — accumulated |diff| inside the closed cycle.
QState.wave_trit_density_pct — share of non-zero trits in the cycle.
9 telemetry getters (parity with the lab fork): getVarianceEMA, getPosThreshold, getNegThreshold, getChargeLimit, getBaselineEMA, getTritRing, getRingHead, setBaselineEMA, setVarianceEMA.
Regression equivalence suite (benchmarks/regression_equivalence.py): 104 096 samples -> 0 state mismatches vs v2.0.
license=GPL-3.0-only in library.properties.
Changed
Static footprint: 84 B RAM (avr-g++ 7.3, ATmega328P); Flash 1758 B.
Docs, keywords, metadata updated.
Verified
100% integer, zero heap, deterministic O(1).
Official suite PASS: AVR compile, O(1) determinism, AAMI EC57 clinical, 4/4 stress tests.