Integral-cover connector mounting checks — 2026-09-26

This supersedes the separate DB9 panel/insert-hanger design. CAD evidence, not physical acceptance. Installed OpenSCAD 2021.01 and system Python/KiCad were used without downloads. PCB files were read only; changes are confined to the case and agent current-state record.

Geometry and outputs

Shell remains 106 × 86.8 × 37.4 mm; carrier underside/top Z7.4/9, brass bodies 10 mm above carrier, module underside Z19, roof underside Z35. RF overlay remains removed; PCB layout/copper/reserve unchanged.

Integral panel: nominal X33…81 (0.02 mm union overlap into each wall), Y−0.47…1.93, Z21.07…35, 2.4 mm clamp thickness. Direct roof connection, two 6 mm nominal forward ribs (inner edges unchanged) outside the flange; no DB9 heat inserts or independent attachment screws.
Closed Ø3.6 post-clearance holes, 25 mm spacing, axis Z26.84762024. No M3/4-40 inference.
Bottom-open 18.078038 mm shell slot; top Z31.99762024. Minimum hole-to-slot web 1.660981 mm; top rail 3.00238 mm. Panel bottom clears module PCB by 0.5 mm; anchors begin 3.10 mm behind the panel rear. Flange registration comes from extracted STEP faces, not the module mounting-hole pitch.
Former 1 mm side slits closed by panel/rib extensions into wall boundaries. Base shoulders: X33.25…80.75, top Z20.82; central tongue width 17.578038, top Z21.69762024 (0.6 below STEP shell). Upstand Y−3.4…−0.5, depth 2.9: PCB-edge clearance 0.5, anchor separation 5.53. Shoulder/edge/tongue seams 0.25. No undercut, no change to 36 mm cable approach width.
Front-wall service recess extends down to lid seam. USB opening stays unchanged; installation is vertical with all cables and both connector posts removed, not a horizontal slide under button guides.
Deleted panel-default.stl, panel mode and layout object. db_panel=false is access-only. Base insert count is 10 M3 + 4 M2.

Affected base/lid exports regenerated; unchanged plungers retained and revalidated under bounded 180-second command timeouts; CGAL reported Simple: yes:

Export	Connected triangle components	Triangles
`base-default.stl`	1	8404
`lid-default.stl`	1	12712
`plungers-default.stl`	3	1716

Base and lid geometry changed only at the connector closure; plunger geometry is unchanged. The lid is one connected printable part, not overlapping separate exports. Three PNGs (assembly-preview.png, db9-preview.png, lid-detail.png) regenerated headlessly under 120-second timeouts and visually reviewed. The bare-lid view exposes the closed post holes, open-bottom shell slot, forward ribs and downward guides; the assembled view shows reseated post-body references outside the panel. Simplified background references are not production component CAD.

Source, expressions and mesh regression

Run from this directory:

/usr/bin/python3 -B validate.py --source --installation --variants

PASS:

Native saved PCB outline, thickness, placements/orientations, carrier/module NPTH centres and diameters, no invented ESP32 mounting holes, button centres and measured USB mouth graphics.
Extracted STEP flange plane/lateral bounds, shell size and true 25 mm lock axes. Read-only KiCad emitted its known enum-property assertions but completed.
Actual SCAD-evaluated stack/connector expressions; independent raised-carrier and taller-module fixtures; stud engagement; obsolete panel/spacer rejection.
Negative assertions for insufficient stud reach, bottoming studs, thin shell/panel, negative flange gap, panel entering the PCB plane, oversized lock holes consuming slot ligaments, ribs entering flange lateral bounds, insufficient sliding seam/shell clearance and inverted closure steps.
Every STL triangle nondegenerate, every edge used twice with opposite winding, expected component counts, positive total volume, print-bed Z and expected bounds. These are edge-manifold checks, not exhaustive vertex-manifold or self-intersection proof.

Continuous installation sweep

--installation renders CGAL intersections of the actual lid with rectangular hardware envelopes swept continuously through 60 mm of relative vertical travel, from below to the final position. A remote 1 mm³ sentinel makes an empty intersection exportable; a passing result has only that cube, one component and exactly 1 mm³ volume.

Included envelopes: maximum shell root/lip, flange, STEP anchor-bracket bounds, RS-232/ESP32/OLED boards, antenna, provisional OLED glass and USB shells, full button-body/actuator bounding boxes and conservative square base-boss bounds. Flange contact is inset 0.01 mm solely to avoid a coincident CGAL face; physical tolerances are not inferred. No cable envelopes are included because cables must be disconnected. Plungers move with the lid and contact the switches at final seating; their return/retention remains an independent physical gate.

PASS default: posts removed, zero modeled swept interference.
PASS variant: carrier_boss_h=7, module_gap=12, lid_above_carrier=28, panel_t=3.
PASS clearance variant: seam_clearance=0.4, shell_margin=0.8, end_clearance=1.5.
Each test also intersects the actual base/lid at seating, sweeps both upstand steps separately downward 60 mm against the lid, and sweeps electronics envelopes upward 60 mm against the new base material (equivalent to vertical electronics insertion). Only the sentinel remains. Existing conservative base-boss envelopes remain included.
An initial full-base Minkowski sweep exceeded the outer 240-second command timeout after default passed. Replacing it with a whole-upstand hull introduced false diagonal interference; the final per-step exact sweep avoids that invented material and completes within the bounded runs.
PASS negative control: restoring original post-body envelopes causes detected interference, demonstrating that leaving posts installed is not an allowed assembly path.

The first thicker-panel sweep exposed a thin residual inner-wall strip where the panel front lay forward of the original wall's inner face. The recess now clears at least the full wall thickness, independently of panel depth; the corrected fixture passes. The recess starts at the lid seam; the new forward-only base upstand occupies its lower portion.

These are selected conservative nominal envelopes, not a full STEP solid-motion/real-component qualification. Unknown socket bodies, component details, wires, tolerances and actual USB/glass heights remain outside the proof.

Parameter CGAL smoke tests

Rendered to temporary files (180-second limit per render), then checked with validate.mesh_check(path, 1) for a single closed, consistently wound edge-manifold component:

QT_QPA_PLATFORM=offscreen openscad -o /tmp/case-integral-variant-lid.stl -D 'part="lid"' -D 'carrier_boss_h=7' -D 'module_gap=12' -D 'lid_above_carrier=28' -D 'panel_t=3' -D 'emboss=false' enclosure.scad
QT_QPA_PLATFORM=offscreen openscad -o /tmp/case-integral-access-lid.stl -D 'part="lid"' -D 'db_panel=false' -D 'emboss=false' enclosure.scad
QT_QPA_PLATFORM=offscreen openscad -o /tmp/case-integral-variant-base.stl -D 'part="base"' -D 'daughter_supports=false' -D 'side_extra=14' -D 'm3_insert_d=4.4' enclosure.scad

The repeatable --variants suite now covers these three configurations plus a raised/thicker-panel base. All four PASS CGAL and one-component edge-manifold checks. These exercise a thicker integral panel with raised stack, access-only lid and wider base/alternate insert pocket. They are not selected or physically validated configurations. Arbitrary parameter combinations are not guaranteed valid. git diff --check also passed.

Hardware gates — unresolved, not hidden by CAD success

Verify posts really unscrew and determine actual thread/pitch, under-shoulder stud length, female depth, required engagement, clamp torque and compatibility with any captive connector hardware. No thread is established by the STEP bore.
Default panel consumes 2.4 mm additional reach: engagement = stud length − 2.4 − washers (− any flange gap). Stock studs may be too short. Verified longer-stud posts/extensions may be needed; do not bottom them or use ordinary screws in the cable-lock mouths.
A same-length 4.8 mm body reseated outside the panel moves its mouth from Y−2.870 to Y−5.270, now 1.4 mm ahead of the shell nose. Full mating may fail despite adequate thread engagement. A compatible shorter-body/panel-mount solution must be measured; a nominal 2.4 mm body would restore the old mouth plane but is not a qualified available part or sufficient female-thread-depth claim.
Preserve full cable mating/locking and external nose protrusion (0.47 mm beyond ordinary wall, 3.4 mm beyond panel); verify actual hood and wrench access inside the 36 mm central space between ribs. Revise geometry/hardware if necessary rather than forcing a fit.
Existing gates remain: insert selection, underside solder/prong and conductive shoulder clearances, header/brass stack agreement, USB envelopes, glass height, button preload/return/overload, material/print orientation and cable-load strength.

No physical print, slicer analysis, insert installation, connector/thread measurement, full cable mating, load, RF, thermal, firmware or device test was performed. The stronger integral load path is designed and modeled, not mechanically load-qualified.

9.4 KiB Raw Blame History Unescape Escape