Comparative survey next to the V-SLOT lead so the trade-off survives
until the FEM stage: industrial T-slot and TIG-welded aluminium for
the frame, heavy-duty telescopic slides or gate-hardware steel
rollers for the motion, and the salvage lead of a truck sliding-roof
mechanism; precision linear guides ruled out for dust and water.
The owner proposes building the roof structure in V-SLOT extrusion,
which fits the house ecosystem and makes the V-groove itself the
sliding rail. Record it as the retained lead with its three guard
rails: vibration-proofed joints, weather-protected rolling interface,
and C-beam/4080 main beams sized by FreeCAD FEM for the deployed
cantilever. Positive mechanical travel lock stays independent.
Official manufacturer figures for the whole energy chain: US3000C
32.0 kg (bank 128), MultiPlus 30.0, MPPT 4.5, Fronius Primo 21.5
(the one estimate that was off, +5.5 kg), Cerbo ~0.4 (reseller value,
absent from Victron docs), LONGi panel 19.5 confirmed. Weight
tracker, architecture, roof concept and the decision entry all
updated; Fronius MPP window corrected to 80-800 V.
The LONGi label photo yields the exact model (LR4-60HPH-370M) and its
electrical figures. The voltages nearly settle the DC/AC split on
their own: ten panels in series reach 409 V, beyond the 250 V MPPT
ceiling but inside the Fronius MPP window, while the MPPT tops out at
five-panel strings. Fixed deck to MPPT as 2x5, sliding deck to
Fronius as 1x10, one conductor pair through the flexible link.
Copyrighted manufacturer PDFs stay local now that the repo is
public; only the versioned INDEX.md with sources and checksums is
tracked so the folder can be rebuilt from upstream.
First numbered estimate for the solar roof: 20 LONGi 370 panels at
about 19.5 kg each plus fixed rails, the self-supporting sliding
frame (aluminium mandatory, steel would double it) and cabling land
around 650-700 kg. Recorded in the weight tracker and the roof
concept doc with the structural and centre-of-gravity implications.
Owner read the minimal kerb weight off the papers: 11000 kg, half a
tonne under the prospectus figure. Update the vehicle sheet and give
the weight tracker its first theoretical budget: about 5.2 t of
fit-out allowance before the real weighing refines it.
Owner read the axle maxima off the vehicle papers: front 6500 kg,
rear 12000 kg (sum 18500 vs 18000 kg PTAC, normal repartition slack).
Replaces the labeled C954E extrapolation — front matched, real rear
is 500 kg higher. Vehicle sheet, weight tracker zero-point and the
research report annotation all updated.
FreeCAD drops timestamped .FCBak backups next to saved documents;
one slipped into the previous commit. Remove it and ignore the
pattern so only the real .FCStd models are versioned.
Owner corrected the variant: it is the 12 m Axer (11.99 x 2.50 m),
not the 12.8 m. Sweep the repo with the researched figures: PTAC
18000 kg (official prospectus, three concordant sources — the
18500/19000 ad-grade values were wrong), PV ~11500 kg for the 12 m,
wheelbase 6270 mm, C954E axle figures kept as labeled extrapolation.
Research report annotated, vehicle sheet and weight tracker updated,
FreeCAD dimension sheet recalibrated (11400 x 2440 x 2010 interior
placeholders).
Owner read the CNIT off the registration card. This code indexes the
French type approval, the most direct path to official PTAC and
per-axle masses; the running research agent was redirected to cross
it against French databases.
Owner confirmed the long-wheelbase 12.8 m Axer. Pin it in the vehicle
sheet with the matching PTAC order of magnitude (~19 t, ad-grade,
official value still to be read on the plate), and bump the FreeCAD
interior-length placeholder from a 12 m guess to 12300 mm so the
parametric zoning reflects the right envelope until the real survey.
Owner confirmed the manual ZF 6S1600 (no Voith automatic), closing
one of the on-vehicle verification items from the research report;
engine row also pinned to Cursor 8 while editing the sheet.
The owner kept the original belted seats at removal time. Record the
reuse decision in the seats sheet (original TCP-approved seats back
on original anchor rails is the strongest RTI scenario) and update
the initial-state inventory; DREAL confirmation item stays open for
the 2003 pre-2005/39-41 nuance.
The owner confirmed the bus was first registered in 2003. Pin it in
the vehicle sheet (C956E / Cursor 8 era, consistent with the Iveco
engine) and turn the blocking DREAL question from generic to dated:
2003 predates the 2005/39-41 seat-belt directives referenced by the
RTI dossier, so original TCP-approved seats need explicit
confirmation.
Deep regulatory research (15+ official sources, RTI 03.8.3 sheet read
in full): the TCP-to-VASP procedure is nailed down (13-item dossier,
seat reduction handled inside it, 86.90 EUR), but two blocking
uncertainties surfaced. First-registration date may forbid the
conversion or require a ministerial derogation, and no firm text
covers the sliding solar deck. Both flagged as blocking questions in
parcours-vasp.md and solaire-toit.md: DREAL first, fabrication later.
Third leg of the CAD tooling: FreeCAD 1.1 for 3D and dimensioned
technical drawings. Seed a spreadsheet-driven model of the interior
(envelope, 9-seat zone, shower cell, Tiroclass bench, fab-machine
zone) so the survey of the emptied bus only has to fill in the
dimension sheet to recalibrate the whole layout. Conventions and
planned files documented in the folder README; decision registered.
Block-diagram the validated architecture as movable shapes and texts:
PV decks feeding MPPT (DC) and Fronius (AC-coupled), MultiPlus and
Pylontech bank on a highlighted 48V DC bus, P17 shore input, AC and
DC boards, Cerbo GX. Gives the electrical folios a reviewable
starting point instead of a blank page.
QElectroTech 0.100.1 is now installed locally, so the qet/ folder
gets its actual project file: five pre-named folios matching the
documented plan (synoptic, 48V DC, 230V AC, rooftop PV including the
sliding deck, monitoring/interlocks) with house title-block fields.
README updated from install instructions to open instructions.
Split the schematic tooling by audience: QElectroTech (IEC 60617,
versionable XML) for the 48V/230V/PV installation folios that
electricians and inspectors read, KiCad kept for in-house boards
(monitoring, solar-deck interlock). Seed qet/ with the folio plan
and conventions; record the decision.
The user picked a proper name for the project repo. Rename on Gitea
(electron-rare/Beaujol-Bus-Lab), flip visibility to public now that
the privacy pass removed face photos and the seller ad screenshot,
and update every reference: local remote URL, CLAUDE.md git section,
blog about page and nav link. Add the missing hero image CSS rule.
The blog needed a visual identity and a path back to the source:
hero photo of the Axer on the home page, village photo and repo link
on the about page, and a git-clone nav link.
Privacy pass before the repo goes public: move the two cab photos
with an identifiable face and the seller ad screenshot out of the
repo (kept in BUS-photos-privees locally), update INDEX.md and
CLAUDE.md accordingly.
Sort and compress 18 site photos into domain folders (EXIF-dated,
indexed), correct battery model to Pylontech US3000C across energy
docs, and publish five backdated posts telling the story so far:
gear-before-bus, arrival, seat removal, sorting, energy architecture,
workbench. Night cab photos with an identifiable face stay out of the
public blog images.
Document the real bus.saillant.cc exposure chain (cloudflared ->
Traefik -> Tower nginx pinned by digest) now that Task 7 hosting
and Task 8 deploy.sh have been validated end-to-end.
Automate publishing the Hugo blog to the nginx host on Tower: build
with hugo --minify, rsync the public/ output to bus-blog, then smoke
test the public HTTPS URL to confirm the route is live end-to-end.
Scaffold the bus.saillant.cc blog: Hugo site config, a hand-written
vendored theme (terminal-bus, zero deps), and the first post plus an
about page.
The theme is minimal on purpose: single baseof/home/page/section
layouts and one CSS file, styled as a terminal/hackerspace prompt
(green-on-dark, ASCII banner, monospace). No upstream theme pulled in,
per the no-blind-dependency policy.
Consumes journal entries later (Task 4); build output feeds deploy.sh
(Task 8). blog/public/ and blog/resources/ stay gitignored.
L'électron rare