Hermetic Normal Calibration¶
Observation card
| Field | Value |
|---|---|
| Status | partial |
| Verified against | docs ontology 5ce15c13; sealed outcome 6e69d792; observer-exclude film 36b003bd |
| SceneId | hermetic_chamber_v0 |
| SceneClass | instrument_calibration |
| InstrumentTier | geometric (+ semantic gate) |
| Channel IDs (now) | cathedral.probe.outcome (gate); Display Modes: NormalRGB / NdotV / Depth |
| Channel IDs (planned) | geometry.normal, geometry.depth, geometry.ndotv |
| Dependencies | Complete outcome plane before interpreting NormalRGB as “hits” |
| Units | normal components dimensionless; depth scene units |
| Validity | HitGeometry for orientation claims; sealed normals if claiming vector truth |
| Display mapping | NormalRGB = f(n); NdotV = f(n·v); Depth = f(distance) |
| Claim boundary | No material nmap; no GR; presentation ≠ probe |
| Last qualified run | not yet qualified |
Why this room is a classroom¶
The hermetic enclosure is for teaching:
- Closure — known wall planes; observer inside a closed box under a stated test.
- Orientation — distinct face normals (floor / walls / ceiling) when hits and normals are valid.
- Policy contrast — straight vs field-on comparison at fixed pose.
- Calibration failure modes — e.g. player self-intersection as a failure (rays hitting the observer avatar) rather than “missing the room.”
It is not a beauty stage and not a wormhole proof.
Status of storyboard layers¶
| Step | Layer | Status |
|---|---|---|
| 1 | Viewport (hermetic box) | implemented (scene) |
| 2 | Completed outcome plane | implemented |
| 3 | Geometry-hit mask (HitGeometry) |
implemented (semantic); UI overlay partial |
| 4 | Raw normal vectors | planned / partial as sealed geometry.normal; host may shade without sealing |
| 5 | NormalRGB mapping | available as Display Mode — not sealed channel |
| 6 | Wall-plane seams in NormalRGB | partial — faceting + resolution + validity |
| 7 | Depth shading | available as Display Mode |
| 8 | NdotV | available as Display Mode |
| 9 | Qualified live multi-face result | not yet qualified |
Current status remains partial until raw normal channels and qualified live multi-face results exist.
Separate four questions¶
| Question | Authority |
|---|---|
| Did transport hit geometry? | Outcome plane (HitGeometry) |
| Is the normal valid? | Per-sample normal validity (when available) |
| How is orientation displayed? | Display Mode (NormalRGB / NdotV) |
| Does the enclosure close? | Closure / six-axis enclosure test—not plate magenta |
If (1) shows HitGeometry but (3) looks wrong, suspect presentation or normal validity, not “no hit.”
Straight vs field-on¶
| Setting | Use |
|---|---|
| Field 0 (STRAIGHT) | Baseline enclosure + orientation classroom |
| Field mid / 1 (FULL) | Same pose; compare semantic histogram and (if justified) display mapping deformation |
Field dial ≠ evaluated field.magnitude channel (planned).
After observer self-intersection correction (36b003bd lineage), expect increased structured room hits when the plate no longer samples the avatar—still confirm with the outcome plane, not color alone.
Storyboard (progressive disclosure)¶
1. Viewport¶
See: structural walls, floor, ceiling; observer inside. Role: world viewport. Why: establish closed geometry.
2. Completed outcome plane¶
See: histogram / class map for full frame. Expect: high HitGeometry on walls/floor if closed and straight/mild field. Gate: lifecycle Complete, Unprocessed = 0. Role: semantic outcome plane (data—not RGB).
3. Geometry-hit mask¶
See: mask where code = HitGeometry.
Not: RGB threshold.
4. Raw normal vectors¶
Status: treat as planned/partial for portable channel. If unavailable: stop geometric-normal success claims.
5. NormalRGB mapping¶
See: Observation Plate under NormalRGB Display Mode. Mapping only. Large flat colors ≈ large planar faces when hits+normals valid.
6. Wall-plane seams¶
See: color discontinuities where face normals jump. Weak seams at 80×45 are expected—not automatic transport failure.
7–8. Depth / NdotV¶
Same hits, different Display Modes—teaching Display ≠ Instrument.
Controls (default)¶
| Control | Setting |
|---|---|
| Display shell | Hermetic |
| Mode | SNAPSHOT Complete |
| Field | 0.00 STRAIGHT baseline, then optional field-on |
| Display Mode cycle | NormalRGB → Depth → NdotV |
| Resolution | 80×45 then 160×90 for evidence stills |
Acceptance (calibration)¶
- Complete plane; HitGeometry count stable under fixed pose.
- Changing only Display Mode changes plate, not outcome histogram.
- No claim of material normal maps.
- If raw normals missing: document unavailable, do not invent.
- Multi-face partition not yet qualified as a public live success until sealed evidence exists.
Failure notebook (presentation-only baseline): Hermetic live failure state.
Interpretation boundary¶
Claim boundary
Collision/face normals ≠ PBR normal maps. Display ≠ probe. NormalRGB does not prove semantic geometry resolution alone. Closure is policy-relative enclosure—not exotic topology. Geometric-normal success requires HitGeometry and trustworthy normals.