Quadratic ear baseline¶
The left-ear master and Ear_Left_Outline library footprint use a quadratic
baseline with short smooth transitions into the existing rounded corners.
This replaces the raster-traced bottom segments only. The two side edges,
rounded corners, overall bounds, both moats, LEDs and connector stay in place.
The fitted quadratic, in left-master PCB millimetres, is
B(t) = (1-t)^2 P0 + 2t(1-t) P1 + t^2 P2, with:
| Control | X (mm) | Y (mm) |
|---|---|---|
| P0 | 114.158073762 | 122.814902750 |
| P1 | 157.835870719 | 107.875016571 |
| P2 | 191.753319509 | 136.807828288 |
The board uses t=0.04 through t=0.96. Short cubic transitions connect that
section to the held corner endpoints at (113.835584,122.616330) and
(192.087029,136.628679), with continuous tangent direction. KiCad represents
the quadratic section as its exactly equivalent degree-elevated cubic, so the
baseline consists of three native Bezier graphics on Edge.Cuts rather than
another hand-traced chain of straight segments.
The control data records the full precision controls,
coordinate frame and pre-change reference. Library coordinates subtract the
outline footprint's (100,70)mm placement. The front 5V and rear GND zone
boundaries follow the same baseline through sampled polygon vertices; their
other relief, clearances and electrical settings are retained. The small rear
5V entry zone and the then-current two deferred touch electrodes were unchanged
by the baseline edit. The later three-electrode revision
supersedes that electrode population without changing this baseline.
Validation, 2026-10-23¶
Independent native saved-file readback measured 0.37905mm maximum symmetric outline deviation from the previous trace. All 169 retained outer line segments are exact; all non-outline footprint poses, numbered pad positions/nets and tracks are unchanged. Both moats and overall bounds are exact, and the library curve controls match the placed instance. Front pads remain at least 1.022mm from the final board boundary.
Before/after DRC on isolated copies with the authored project rules and matching local libraries gives the same findings: 199 silk-over-copper warnings, 14 existing library mismatches and one deferred touch unconnected item. No new copper, edge, width or connectivity finding was introduced. Both modified rail zones were refilled. This is geometry/electrical-CAD validation, not mounted fit, bend, thermal or supplier acceptance.
The preserved study and verification outputs are under
build/sensing-study/2026-10-23/baseline-only/; ordinary build/ cleanup can
remove them. The native source curves and this control data retain the design.
The frozen historical fit assets are not refreshed by this change.