Ambient false alarm held
0.0 of 10% target
Live modal growth rate σ
Each point is the monitor’s causal window score as the stream advances.
The alarm fires where σ crosses the certified threshold; the onset marker is the
independently estimated oscillation onset.
In-band envelope (context)
Mean analytic envelope of the clean frequency channels around the documented
1.13 Hz mode — the physical growth the detector reads.
Alarm log (table view)
| # | Stream time (s) | σ (1/s) | vs onset |
|---|
The sealed chain
Every link is content-addressed. Anyone with the cited public capture can recompute
all three hashes and replay this page’s claim.
Raw capture (citation-only, UTK library)
f30d399b017aeee04987dadbf9cec0117444b4372456c93aa367e86ccba3772fCertified operating point — sealed cross-dataset evaluation
This replay’s sealed record
aafa496766ab9af6… (regenerable byte-for-byte from the capture + evidence)
Claim boundary (read before quoting)
- The threshold was calibrated on this system’s own pre-onset ambient windows at a
10% false-alarm target — per-system calibration is the deployment step, not a fallback.
- The calibration corpus is small (six ambient windows, three events, p = 0.33):
this is a demonstration of the sealed live pipeline, not a statistical
generalisation claim. The powered leg (WECC 240-bus) is in progress.
- The sentinel is review-only: it observes and seals; it never actuates.