Map a per-channel sensitivity ∈ [0, 1] and the channel's pool to a DetectThresholds. Pure; the server's only source of a tracker's gates. Two gate classes:
CLASS A (latency/prominence) relaxes in BOTH pools as s→1 — music-safe, because the
Class-B discriminators still stand, so admitting a quieter/faster ring never re-opens
a music FP: persistFrames, subCleanFrames (the streak COUNT, not the pshr bar),
paprDb, phprDb, growthMinDbPerFrame.
CLASS B (the gates that actually reject PLAYED MUSIC — #106 single-frame-share, the
#106 crescendo tiebreak, and the total-rise bar) relaxes ONLY in the FIXED
(soundcheck, no-music) pool; in the LIVE (show) pool they stay pinned at today's value
at EVERY sensitivity. pnprDb (lone-tone core), pshrMinDb (#116 sub-harmonic BAR),
and growthMaxDipDb are PINNED in both pools always.
The #140 SATURATED-PLATEAU route arms (plateauFloorDb defined) only at s > 0.
ANCHOR INVARIANT (load-bearing, pinned by a hard test): the s = 0 output for BOTH pools
deep-equals DEFAULT_THRESHOLDS. Note the LIVE-pool safety by-product: with
growthMaxSingleFrameShare pinned at 0.5, a 2-frame step can never confirm (one hop is
the whole rise), so live-pool ramp confirmation still rests on ≥ 3 frames — the
sub-second win there comes from the shorter sub-clean streak, relaxed prominence, and the
plateau route, not from persistFrames → 2.
Map a per-channel
sensitivity∈ [0, 1] and the channel's pool to a DetectThresholds. Pure; the server's only source of a tracker's gates. Two gate classes:s→1— music-safe, because the Class-B discriminators still stand, so admitting a quieter/faster ring never re-opens a music FP:persistFrames,subCleanFrames(the streak COUNT, not the pshr bar),paprDb,phprDb,growthMinDbPerFrame.pnprDb(lone-tone core),pshrMinDb(#116 sub-harmonic BAR), andgrowthMaxDipDbare PINNED in both pools always.The #140 SATURATED-PLATEAU route arms (
plateauFloorDbdefined) only ats > 0.ANCHOR INVARIANT (load-bearing, pinned by a hard test): the
s = 0output for BOTH pools deep-equals DEFAULT_THRESHOLDS. Note the LIVE-pool safety by-product: withgrowthMaxSingleFrameSharepinned at 0.5, a 2-frame step can never confirm (one hop is the whole rise), so live-pool ramp confirmation still rests on ≥ 3 frames — the sub-second win there comes from the shorter sub-clean streak, relaxed prominence, and the plateau route, not frompersistFrames → 2.