Key findings
Legend: how sure each number is
documented manufacturer specification measured calibrated multi-view photogrammetry, residuals reported estimated single-view estimate or engineering inference with a stated range unknown envelope only: contents not measured proposal cradle design keep-out adjustable reservation
No manufacturer CAD file, dimensioned drawing or teardown was found. Internal parts (PCBs, contacts, key pivots, mounts) are shown only as labelled, conservative envelopes. The underside screw-like recesses come from one photo. Their thread size and length are unknown, and they must not be used as mounting points.
Interactive model
Model functions: buildOtterchipSE25(THREE, options) in otterchip-component-se25.js. Metadata is in group.userData, and part provenance is in mesh.userData.status.
Photo ↔ model comparison (matched calibrated cameras)
Residual and agreement metrics
The fitted points are white-key front gaps, black-key front feet and top edges, button glyph centres, LEDs and the USB receptacle. Silhouette and edge lines are fitted with a support-function residual on a filleted rounded-box case. Withheld points come from a separate fit that excluded them: 3 gaps per view, plus 2 black keys from different views and 2 buttons in the hero view. Mask IoU and chamfer distances compare thresholded photo regions with flat-colour renders of the JavaScript model. The JavaScript model is only an approximation of the fitted parameters. For example, the shell uses assumed walls, and the key fronts are not rounded.
Camera calibration, perspective and lens uncertainty
Dimensions (with ranges)
Height is unresolved: the photo fit and the published 21 mm specification disagree. The model does not establish which datum the manufacturer used. Treat the reported height ranges as fit sensitivity, not manufacturing tolerances.
Ranges are the min–max across 13 calibration variants, including two free-focal multistart solutions of near-equal cost. The variants include the feet-plane gauge at 0.5, 1 and 2 mm, hero focal length fixed at ±20%, each view excluded in turn, and runs with and without fillet priors. The "formal" σ comes from the Jacobian. Systematic spread dominates.