etr: fit RESTORE_GAIN — restoring force green (14 PASS / 0 FAIL / 2 PEND)

Spring restoring force (gain 0.5) toward band centre 26, active only out
of band; pulls the basin into [17,35] in a couple of steps without
overshoot. Closes the two red tests (L3 magnitude, L2 convergence).
NOTE: superseded design incoming — Anja's 4-zone bistable axis (snap
<7 / soft-pull 7-17 / stable / incoherency 35-45 / snap >45) will fold
this in as the soft-pull gain and revise L7.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015hmgREHNsxYCuim33yUF2c

docs/plans/A8-etr.md

@@ -20,8 +20,9 @@ engine, and *not* migration. The axes:
 Handles the stress PS+/Eth-Int generate.
 
 ## 2. Status / certainty
-**SCAFFOLD — built and tested this session** (`run_etr_tests.sh` → 12 PASS / 2 FAIL-by-design /
+**SCAFFOLD — `RESTORE_GAIN` fitted** (`run_etr_tests.sh` → **14 PASS / 0 FAIL / 2 PEND**; the 2 PEND
-2 PEND). Full law + status in `../../src/endocrine/etr/etr_invariants.md` (this spec defers to it).
+are L5 active coupling and L7). Full law + status in `../../src/endocrine/etr/etr_invariants.md`
+(this spec defers to it).
 
 ## 3. Language & location
 GNU Octave · `src/endocrine/etr/` (`etr.m`, `test_etr.m`, `run_etr_tests.sh`, `etr_invariants.md`).

src/endocrine/etr/etr.m

@@ -22,6 +22,12 @@ function v = ETR_BAND_LO();  v = 17; endfunction
 function v = ETR_BAND_HI();  v = 35; endfunction
 function v = ETR_WRAP();     v = 50; endfunction
 
+% ---- fitted constant (NOT law) ----
+% RESTORE_GAIN: spring stiffness of the out-of-band restoring force (etr_axis_restoring).
+% Fitted to satisfy L3-magnitude (force nonzero out of band) and L2 (zero-drift
+% convergence into the band) without overshoot past the far edge. 0 < GAIN < ~2.
+function v = ETR_RESTORE_GAIN(); v = 0.5; endfunction
+
 % ---- state: one point, three scalars ----
 function s = etr_init(coord)
   if nargin < 1, coord = [25 25 25]; endif   % default: a valid in-band point
@@ -50,11 +56,22 @@ function d = etr_axis_restoring_dir(v)
   endif
 endfunction
 
-% ---- L3: per-axis restoring FORCE (STUB: magnitude UNFITTED -> RED) ----
+% ---- L3: per-axis restoring FORCE (FITTED) ----
-% Direction above is law; the magnitude/curve (RESTORE_GAIN) is C1 TBD.
+% Direction is law (etr_axis_restoring_dir); the magnitude is a spring fitted
-% Returns 0 for now so the magnitude + convergence tests stay red.
+% via RESTORE_GAIN. The force is active ONLY out of band (in band = slack, L3),
+% and pulls toward the band CENTRE (not the near edge): a centre target makes the
+% step cross INTO [17,35] and stop there, whereas an edge target would asymptote
+% onto 17 from below and never reach |v|>=17 (failing L2). Below-band pushes away
+% from 0, so the restoring force alone never crosses zero (L7).
 function r = etr_axis_restoring(v)
-  r = 0;  % STUB — TODO: RESTORE_GAIN * shape(v) along etr_axis_restoring_dir(v)
+  GAIN = ETR_RESTORE_GAIN();
+  a = abs(v);
+  if a < ETR_BAND_LO() || a > ETR_BAND_HI()
+    centre = (ETR_BAND_LO() + ETR_BAND_HI()) / 2;   % 26
+    r = GAIN * (centre * sign(v) - v);              % spring toward signed centre
+  else
+    r = 0;                                          % in band: slack (L3)
+  endif
 endfunction
 
 % ---- L5: cross-axis coupling (OPEN SEAM) ----

src/endocrine/etr/etr_invariants.md

@@ -24,10 +24,15 @@ Each axis is a standalone scalar carrying one of ETR's three tensions:
 | L7 no-zero-cross | **C2** | Bands wall off 0; a pole/sign flip happens **only** by riding over the ±50 wrap, never through neutrality. *(emergent — unconfirmed)* |
 | L8 mechanism | **C2** | "Opposition" = a restoring force on the drift, **not** an out-of-band cost. *(unconfirmed)* |
 
-## Constants (NOT yet fitted — C1)
+## Constants
 `BAND_LO = 17`, `BAND_HI = 35`, `WRAP = 50` are **law** (L1–L3), not fitted.
-`RESTORE_GAIN` (restoring magnitude/curve) and `COUPLING` (L5 strength/mapping) are **TBD**,
+`RESTORE_GAIN = 0.5` is **FITTED** — the spring stiffness of the out-of-band restoring force: active
-to be fitted so the tests pass — never asserted ahead of a test.
+only out of band (in band is slack, L3), pulling toward the band **centre (26)** so a step crosses
+*into* `[17,35]` and stops there (an edge target would asymptote onto 17 from below and fail L2).
+Below-band pushes away from 0, so the force alone never crosses zero (L7). Lands e.g. `[5 5 5] → 20.75`
+in 2 steps; `[45 −45 40] → [30.75 −30.75 33]`. Valid range `0 < GAIN < ~2` (above ~2 a cross-in step can
+overshoot the far edge).
+`COUPLING` (L5 strength/mapping) remains **TBD** — open seam, not to be invented.
 
 ## Testable predicates (see `test_etr.m`)
 - **L1**: `wrap(50) = −50`; `wrap(60) = −40`; `wrap(v)=v` for `v∈(−50,50)`; `wrap(49.9) = wrap(−50.1)`.
@@ -58,7 +63,9 @@ scaffold seam is correct). What's fixed is their **provenance** (upstream in SAE
 endocrines) and two side-effects (reshuffle, shift-rate) that belong to *those* organs.
 Still open: which stress endomotiv; the "too strong" reshuffle threshold; the shift-rate function.
 
-## Status at scaffold
+## Status (RESTORE_GAIN fitted — 14 PASS / 0 FAIL / 2 PEND)
-Implemented (green): L1 wrap, L3 direction, L4 drift-must-be-fed, L5 coupling-off identity.
+Green: L1 wrap, L3 direction + **restoring magnitude (fitted)**, **L2 band-convergence**,
-Red (await fitting): L3 restoring **magnitude**, L2 band-convergence.
+L4 drift-must-be-fed, L5 coupling-off identity. L8 (opposition = a restoring force on the drift, not an
-Pending (unconfirmed): L5 active coupling, L7 no-zero-crossing, L8 mechanism.
+out-of-band cost) is now **realised by construction** — the force is added alongside drift, never charged
+as a cost — though still tagged C2 until a dedicated test pins it.
+Pending: L5 active coupling (C1, open seam), L7 no-zero-crossing (C2).