docs: NOTES — verified comb params (neg sens, multi-mode); comb needs combine/acc, not params

handoff/NOTES_LEVEL.md

@@ -1137,3 +1137,29 @@ band sens 2.24 for two of four bands, real reducer output is ~neutral, ours stil
 Likely the bridge LUT/level_scale is calibrated on the al_* fit levels, not on this
 low/wide band config. Same class as the known low-level over-cut (al lv18/24, NOTES:964).
 Next: calibrate level_scale per input level / verify sens enters the level path correctly.
+
+## ============ UPDATE 2026-08-21: COMB PARAMS VERIFIED — two SEPARATE phenomena ============
+Decoded true comb wav params from comb_all.rpp (real 4-band config, NOT what tests assumed):
+  band1 fc=1000 q=1.0  sens=+12 mode 1
+  band2 fc=1778.7 q=4.5 sens=-12 mode 1   <-- NEGATIVE sens
+  band3 fc=195.1 q=0.494 sens=+2.24 mode 0
+  band4 fc=13408 q=1.0  sens=+12 mode 3
+  depth=0.864, mode=1, selectivity=10, sharpness=10, resolution=1
+Earlier comb tests fed wrong params (band1=678.76, band2/3 sens=+2.24, band4 q3) — those
+were NOT the reference config. Even with corrected fc/q/sens the model still over-cuts -7..-15 dB.
+
+### Two SEPARATE problems (verified):
+1. band1-ONLY (fc=1000 q1 s12, input comb.wav -25.6 dBFS): real soothe2 cuts tones ~5.9 dB
+   @1000 (comb_b1_1000 real ref, NOT neutral); our model cuts -8..-15 (out-real = -2.8..-7.7).
+   => BASELINE depth is too big at low level (level-scale/LUT low-tail over-cut).
+2. comb_ALL (=comb_ref≈comb_byp, reduce ~0 dB on all tones): real soothe2 returns to NEUTRAL
+   despite band1 active. => MULTIBAND interplay (combine/acc 0x5407c8) must cancel band1's cut,
+   consistent with M11 additive red=base+ΣW·H and band2 sens=-12 acting as a positive adder.
+   This is the combine gap (NOTES:1071 F2), and it needs negative-sens + per-band mode handling.
+
+### Model gaps surfaced:
+- negative sens (band2=-12) not modeled -> band2 treated as +suppressor instead of detractor.
+- multiband min() (not additive combine) -> cannot reproduce the all-band cancellation to ~0.
+- per-band mode 0/3 (peak vs shelf/tilt band types, M10) not dispatched.
+No code change; investigation only. These require the structural combine/acc + per-band type
+before 4-band comb can close.