Completed: imported the exact HCTL HC-RJ45-5JA-4-2 manufacturer STEP, attached under the existing footprint, and verified its orientation from below and side. No parametric approximation was required. This is model-placement completion, not electrical, manufacturing, or connector-fit certification.

Manufacturer source: https://hctldz.com/product-3-28/1025.html ; downloadable archive: https://hctldz.com/static/upload/2026/10/14/202610145047.zip . Archive contains the exact-part STEP, dated 2025-12-08. Original source SHA-256: da0ae6f7b063ac2aab415e40865e69013f33c87bd14ce8d09a3efc844db90770. Geometry was not modified. Vendor model is a single uncolored solid, so it displays monochrome; LED windows and four ports are geometrically represented, not colored yellow/green.

Mechanical numbers in the last six rows are measured from the exact vendor CAD, not inferred from footprint pad diameters. Nominal CAD is not a tolerance or retention-force specification.

Compared the source STEP section below seating (source Y=-8 mm) with all 52 numbered footprint pad locations, converting KiCad Y-down coordinates to the computed Flux Y-up frame. Matching is one-to-one: 52 distinct terminal coordinates matched to 52 pads. Two mounting axes were separately checked against the existing mounting holes. CAD calculations used an exact STEP import, not a rebuilt approximate model.

Rear shield sections center at X=±7.9, Y=3.86997 mm; footprint holes are (-7.88,3.78) and (+7.87,3.78) mm. Side shield sections center at X=±31.775, Y=-2.89003 mm; holes are X=±31.75, Y=-2.89 mm. These small vendor-CAD versus footprint differences were retained, not hidden by moving individual model pins or editing the footprint. Terminal section envelopes remain within their respective nominal hole apertures. Split mounting snaps intentionally have retention geometry; no elastic insertion-force claim is made.

Warning — preexisting drawing discrepancy: footprint signal drills are approximately Ø1.2 mm for pads 1–32, whereas the supplied drawing calls for Ø1.0 mm. This did not block visualization, but it is not an exact drawing match; the footprint was deliberately left unchanged.

Native 3D readback included both footprint and attached model, 1024-pixel orthographic inspect views:

Baseline commit #4b2b5936, with complete original node-tree/footprint asset readback and all 54 computed pad/mount positions inspected before model placement. Existing footprint asset UID e76d76ac-c816-443a-be3b-9bbf515d43e4 retained the same name and immutable source URL before and after. Source SHA-256: 87e627fb2e2a6b3bca6462b58fb3613b42d074d7d5277690d2adede54a4eac51.

Read-only change-diff inspection proves the only design changes through model placement were:

No pad, mounting hole, symbol, pin, name, terminal-order, net, preexisting rule, or electrical property path changed. This notes document is the only additional documentation change.

Post-mutation automated board DRC validators returned NOT CHECKED for this standalone part project; no clean DRC result is claimed. Model geometry measurement, one-to-one numerical pad matching, direct native 3D readback, and commit-diff preservation checks were completed.

Follow-up checked the exact vendor STEP planar faces, not the overall bounding-box bottom (which includes the through pins). Current native readback confirms footprint origin (0,0,0), model position (0,-3.882,7.1934) mm, rotation (90,-90,0) degrees, and unit scale. The nominal transform is therefore Z = source Y + 7.1934 mm, with board-top seating reference Z=0.

Face indices are zero-based CadQuery/OpenCascade face indices from importing the unchanged STEP; they are reproducible with the source hash recorded above, not platform node IDs. Supporting lip 328 spans world X=-30.425…+30.425 mm and Y=-11.58003002…-8.63003002 mm. The two small support faces lie at X=+29.675…+30.275 and -30.275…-29.675 mm, with Y=-4.82482936…-0.92003002 mm. All three support normals transform to -Z. Their combined area is 184.16937276 mm².

Conclusion: the physical support faces are seated to within 0.0000422 mm (0.0422 µm), the rounding residue of the existing Z offset. The main body has a real 0.300 mm geometric standoff relative to these supports; other underside regions are recessed 0.800 mm. Lowering the model by either visible body clearance would bury the actual support lip into the board. No placement correction is justified, and the model transform and footprint remain unchanged.

The native scene manifest adds 0.035 mm to the model node's stored Z, yielding rendered model translation Z=7.2284 mm; the generated board GLB contains a 0.035 mm outer-copper thickness. Relative to that copper-top convention, applying the manifest's rounded Euler angles gives support elevations +0.0000139…+0.0000531 mm, still effectively zero. The 0.035 mm scene translation must not be interpreted as a floating gap or subtracted again from the node position.

A fresh native right-side readback with footprint included (1024 pixels, zoom 0.5) was inspected after these measurements; no model mutation occurred. That side preview still visually separates the copper line from some housing edges and does not by itself prove literal pixel-level support contact. The generated standalone-footprint scene declares substrateThickness=0 although its board GLB includes a 1.6 mm preview substrate; a preview-datum/rendering discrepancy is possible and was not independently resolved. The seating conclusion above is established by the exact source support faces and actual node transform against the requested Z=0 datum, not by asserting that this ambiguous preview is a calibrated mechanical section.

Reproduction files retained in the stopped package-model computer: /root/work/rj45/seating.py, seating-planes.json, seating_render.py, scene.json, and the unchanged source STEP at /root/out/HC-RJ45-5JA-4-2-vendor.step. This follow-up changed documentation only.