The neck: head mount and pan/tilt linkages against the stock design (2026-10-31, #133)
Question (#133): in Stage 2/3 the head looked wrong on the neck, the head seemed to turn only about half the pan servo angle and a third of tilt, and the pan four-bar could not close at its lower limit. Re-derive the stem seat (STEM_AT, STEM_IN) and the pan/tilt linkages (PIVOTS, BARS, horns) from the stock design, since the neck is not built yet.
Sources (local, gitignored, docs/references):
- STL:
02-pan-tilt/models(upright v2, tilt arm, pan base, servo mounts, both bars),03-mouth/models/main bracket v2.stl,01-eyes/models/main bracket.stl; trimesh plane sections (circle and loop fits), 0.5 mm steps. - PT-pN: the pan-tilt guide PDF, page N. GIF-PT:
02-pan-tilt/picture-01.gif(72 frames, fixed camera in front of the face about 17 deg up). GIF-M:03-mouth/picture-01.gif. - CONFIG:
02-pan-tilt/remote-config-pan-and-tilt-v1.json: tilt 20..150, rotate (pan) 40..160, start 90.
Findings
Section titled “Findings”- Horn: the kit’s short single arm, 6 holes, bar screw at the second hole from the end: PT-p11 step 3 (“drill a 2mm hole … at the second from the end position”), PT-p12 photo (hole 5 from the hub), PT-p18 step 12 (M2 screw and nut through the smaller bar hole) [M]. Our 12.75 mm (HORN_STEP pin) stands.
- Bars: pan 15, tilt 20: PT-p18 step 12 (“one is now 5mm longer, this is used for the tilt”), p.19 steps 14/15
(pan = the shorter); STL
PandT barholes 15.0 apart,tilt bar20.0 [M]. BARS stand. - Levers (STL, circle fits): tilt arm lever hole r 1.1 along X at (y 26.5, z 33.4), ears r 1.35 at (y -13, z 28.4), both +5 on the raised upright; upright pan lever hole r 1.1 along Z at (-5.5, 25.5), tilt pivot r 1.1 at (y -13, z 33.4); pan base tower holes along X at y -5.0 / 22.5, z 18.4 [M]. Our printed parts carry the same holes. PIVOTS and PT_LEVER_HOLES stand.
- Horn directions at neutral: tilt horn horizontal, bar vertical (PT-p20 step 15, PT-p21 photo); pan horn about perpendicular to its bar (PT-p20 step 14, PT-p21 step 16). Both as the model has them (PT_HORN_DIR).
- Stem seat:
main bracket v2stem 12 x 5 (x 23.5..35.5, z 13..18) runs y -45.6..-40.6 below the rail (y -40.6..-30.6); the tilt arm’s window is x -6..6, y -15.5..-10.5 through its 5 mm plate (z 30.9..35.9 in the STL, +5 raised). The stem fills the window, the rail lies on the arm top: stem bottom (0, -13, 35.9), rail underside z 40.9, eye centres z 81.5, 78.5 above the plate top. STEM_AT stands; its depth goes from inferred to measured. GIF-PT frame 5 and PT-p26: the stem disappears into the window, nothing below, no gap. - Facing: the eyes face -Y, the plate’s pointed end, away from the pack, the boards and the switch (PT-p23, PT-p26). From the front the pan servo sits behind the pan base on the viewer’s left, the tilt servo on the right (GIF-PT frame 5). As modelled.
- Head swing, GIF-PT (pinhole camera fitted over all 72 frames to the eye shells and the pan base outline, residual 1.1 px on the eyes, 1.7 px on the base): pan +19 / -12, tilt -19 (up) / +10 (down), each +-3 deg. Neither horn is in view, and nothing shows the stick at full throw, so these are what the mechanism at least does. GIF-M pans 46 +-6 deg end to end (eye spacing against shell size), the fuller throw.
- The model (stage 2, this geometry): pan -40 -> +24.9, +70 -> -21.4 (46.3 end to end); tilt -60 -> -17.2, +60 -> +16.2. Both GIFs are inside that. The “0.47 x pan, 0.32 x tilt” gains are the stock linkages’ (crank 12.75 against lever radii 26.1 and 39.8), not a model error: at the config’s full throw GIF-PT’s swing would be a ratio of only about 0.25.
- The pan four-bar (crank 12.75, coupler 15, lever 26.09, ground 43.3) is non-Grashof: it closes for horn -43.4..+90 and toggles at -43.4 (bar and lever in line, transmission angle 0). Transmission angle: 78 deg at 0, 36 at -35, 23 at -40, 8 at -43. No hole, bar or horn direction from the sources closes -50..+70 (enumerated: shaft at either tab end, every horn hole 6.75..14.75, either bar, both assembly modes).
Conflicts
Section titled “Conflicts”- Pan range: CONFIG rotate 40 (horn -50) against the stock linkage, which cannot close past -43.4. Linkage (STL) wins: LIMITS rotate 50..160 (horn -40..+70). -40 keeps a 23 deg transmission angle and gives GIF-M’s 46 deg end to end; the textbook 40 deg minimum (about -32) would give 39, below GIF-M. The real Mech at 40 would drive the bar through toggle or stall there.
- Head height: GIF-PT frame 5 reads the eye centres 64..71 above the plate top, the model 78.5. GIF-PT predates the mouth and the 5 mm raise: with the v1 bracket (stem bottom y -40.6, no rail, feet on the arm top) and the v1 upright the same STL stack gives 68.5, inside the reading. STLs (v2 bracket, v2 upright) win; 78.5 stands.
- Eye centres above the arm top: GIF-PT 31..33, v1 STL stack 35.6. One frame, camera pitch assumed 10..25 deg and about 4 % scale: within the reading’s error; STL wins.
- Pan side by side: GIF-PT left pan -10 from the base outline against -13 from the eyes (same frames). Both inside the +-3 the test allows.
- Bars in the PT-p12 photo look equal; PT-p18 text says the tilt bar is 5 mm longer, and the STLs agree with the text.
- Open: which servo sense turns the head which way (the GIFs show no horn). The model takes servo angle - 90 as the horn’s right-handed turn about +Z (model/sim.py joint_deg). If the real pan servo turns the other way, the closable range in servo terms is -90..+43 and the config’s +70 end is the one that toggles; the calliper-free check is to run the built neck to both ends. The swing test compares the sides by size for that reason.
noggin v0.2.26-6-g09bf9b7 · 09bf9b7