# 3D height study with shallow grading and explicit floor thickness

Completed 26 September 2026. Requested update to the height study. All four lifts now have matching 3D geometry, sections, basement elevations, dimensions and interior perspectives for two ground settings and two floor build-ups. Existing room layouts and roof geometry are retained.

## Result

Window heights below are geometric whole-frame envelopes before installation allowance, not clear glass dimensions or ordering sizes. Width remains 1.60 m. Minor sub-millimetre calculation differences are rounded for presentation.

| Lift | Natural ground, 30 cm floor | Graded, 30 cm floor | Natural ground, 33 cm floor | Graded, 33 cm floor |
| --- | --- | --- | --- | --- |
| 0 cm | 25 / 26 cm | 30 / 30 cm | 22 / 23 cm | 27 / 27 cm |
| +20 cm | 45 / 46 cm | 50 / 50 cm | 42 / 43 cm | 47 / 47 cm |
| +30 cm | 55 / 56 cm | 60 / 60 cm | 52 / 53 cm | 57 / 57 cm |
| +40 cm | 65 / 66 cm | 70 / 70 cm | 62 / 63 cm | 67 / 67 cm |

Default is shallow grading and the 33 cm assembly reserve. Controls reload the matching complete dataset; lift buttons still animate using the main page's renderer without moving the outside camera.

## Datum is separate from excavation

The owner measured 30 cm over the existing frame and estimated approximately 10 cm below it to ground. A 10 cm distance from frame top to the old ceiling is a **working assumption**, supported only qualitatively by interior photographs showing a reveal below the ceiling. Existing kitchen frames are placed at −0.40 to −0.70 in the study, with a ceiling at −0.30. Ground at the existing reference is therefore −0.80. The two DTM reference strips differ by about 1.4 cm, so the old frames have roughly 9.3 and 10.7 cm ground clearance with the averaged datum.

The previous DTM reference averaged −0.707925. The new study shifts the DTM datum down 0.092075 m. **This is a correction of the assumed house-to-terrain relationship, not 9.2 cm of proposed earth removal.** It is displayed separately from the grading quantity. The absolute height connection is not surveyed. A different actual old head clearance or floor-to-ceiling elevation changes this working datum. Holding the new 10 cm head reserve fixed, each 5 cm change in the assumed old head gap changes predicted new window height by 5 cm.

The relative DTM rise between old windows and new windows remains 5.06 / 3.86 cm without grading. The source one-metre terrain grid has not become centimetre-accurate through interpolation. The coarse photographic reconstruction has no independently calibrated scale or ground-level error bound and cannot remove this limitation.

Source review and relative derivation: `planning/height-datum-41/README.md`, `planning/reset/evidence/owner-window-measurement-40.md`. Canonical `modell.json`, original DTM grid, selected upstairs floor plans and prior application packages are not overwritten by this study.

## Grading geometry and quantities

The proposed field is beside the river-facing wall, x = 3.85645, across y = 9.50 to 13.60 with 0.80 m end blends. At the wall the proposed ground is −0.80. The cap falls 1:50 for the first 3 m, then ties into the lower natural ground. The operation only cuts, never fills. It adds no ground-cutout polygon or enclosed lightwell. A refined 0.25 m display mesh represents the shallow shape; material surfaces are raised only 1–2 mm to avoid z-fighting, not as construction layers.

Numerical integration on a 5 cm grid gives approximately **0.361 m³** removed over **19.84 m²**, with **5.05 cm maximum cut**. This is in-place model volume, not a measured excavation quantity or loose spoil volume. The tan field recolours existing surface triangles whose centres exceed a 1 mm cut threshold, avoiding a separate floating overlay. No terrain outside the graded extent is altered relative to the calibrated natural setting. Sampled outward profiles across the window strip have no upslope; the outer and end edges meet original terrain continuously. Local grading also affects part of the modeled yard surface, so this is not simply two small holes immediately under the windows.

[DiBK TEK17 § 13-11](https://www.dibk.no/regelverk/byggteknisk-forskrift-tek17/13/vi/13-11), checked 25 September 2026, requires surface water to be led away from the building. Its guidance gives 1:50 over 3 m as a preaccepted solution and allows other measures where appropriate. The modeled shape illustrates that principle but does not verify drainage capacity, soil, existing drainage, snow conditions, boundary effects or construction feasibility.

## Layers and vertical chain

The project construction note `soknad/astra/ASTRA-OPPGAVER.md`, section (e), proposes 14 mm parquet + 10 mm impact-sound board + 22 mm chipboard + 198–223 mm joists + 25 mm acoustic suspension + two 12.5 mm plasterboard layers. The arithmetic is **294–319 mm**, not an exact 300 mm in every case. A 330 mm envelope leaves 11 mm over the deeper candidate assembly. These are unverified assembly proposals, not beam sizing or tested acoustic/fire performance. The assumed support line is essential; no 7 m unsupported span has been validated.

- Basement floor −2.75 + lift, interpreted as the study's finished floor reference.
- Main floor 0.00 + lift.
- Ceiling underside −0.30 or −0.33 + lift, matching the actual 3D slab underside and section thickness.
- Clear basement height 2.45 or 2.42 m. Raising the whole extension leaves this unchanged.
- New window top 10 cm below that ceiling, so −0.40 or −0.43 + lift. This separate reserve is not a sized lintel.
- Window bottom is 10 cm above the highest sampled ground under its full width and near-wall strip. Frame render geometry now fits inside the stated whole-frame envelope; the previous decorative frame extended beyond it.
- Upper floor +2.65 with 25 cm floor zone gives the main room its 2.40 m clear height. Loft slab also retains its 25 cm study envelope. The fixed roof and all upstairs headroom calculations are unchanged.
- Basement slab is shown as a 25 cm schematic solid, aligned between section and 3D. It is **not** a complete foundation/insulation/drainage build-up. Excavation depth is not established by that display solid. Adding finishes above the basement floor reference would consume room height unless the base is lowered accordingly.

The 33 cm setting consumes the 3 cm previously omitted from the thicker assembly option. Further ceiling services or a deeper beam consume additional height; the study does not hide them inside a nominal 2.45 m room.

## Verification

- `check.py` verifies all 16 combinations: room polygons and upstairs partitions unchanged, roof/upstairs meshes match prior same-lift meshes, ceiling and slab underside agree, window heads and whole-width ground clearances agree, all section/interior data match 3D, and canonical model hash is unchanged.
- Grading checks cover no fill, continuous boundary ties, outward profiles and quantity calculation. Layer arithmetic is checked independently.
- `browser.mjs` exercises four lifts, both ground controls and both assemblies, desktop/mobile layout, interior perspectives and JavaScript errors. Outside and inside review cameras match saved before coordinates.
- Before images and camera coordinates are under `before/`. After images include close river facade, basement eye level and whole yard, plus every terrain/assembly combination at 0 and +40 cm. Reviewed windows, corner trim, floor junctions, roof, ground support and terrain occlusion. The material field is an intentional grading annotation, not a measured material specification.
- The interior view remains a simple geometric view with schematic lighting and stair rendering. It does not establish daylight levels. Global datum adjustment changes the visible foundation exposure and contextual building bases, as intended for this conditional study.

Build: `npm run build:height-study`. Check: `npm run test:height-study`. Study URL: `/nettside/hoydestudie.html?height=0&ground=trim&stack=33#husmodell`.

Shared production checks: `npm test` passed all seven checks, including bundle/source agreement, solar, stairs, delivery, site, materials and application consistency. These cover the canonical site as well as guarding against shared-renderer regressions; they do not validate the study assumptions as surveyed facts.