CABLE-CHECK, power-route re-verification after the 25 mm head drop
Task #18, 2026-07-07. Geometry probed with trimesh signed-distance / containment
against scenes built as OUT=_cable_t{θ}.glb PAN=0 TILT={-30,-15,0,15,30} (probe
scripts in the session scratchpad; web/assembly.glb untouched, probe GLBs deleted
after the run). Conduit spec: Ø5 virtual conduit (≥2.5 mm clearance to every solid);
real bundle is the Pi power pair, ~Ø3.6.
Route under test: base rear USB-C wall port → base cavity (pan service loop) →
16×8 deck pass at cable_exit (12,−24) → platform slot → neck channel
(neck_chan_y=−26) → column top → head bottom-rear slot (tilt drape) → Pi USB-C
(bottom edge, z≈126).
Verdict summary
| # | Segment | Verdict | Worst number |
|---|---|---|---|
| 1 | USB wall port → cavity | CLEAR | 4.00 mm |
| 2 | Service loop (cavity annulus) | CLEAR (at z 36–45.5) | r40–56 ring free all-round |
| 3 | Deck pass, through the deck itself (z 46–51) | CLEAR | 16.0 × 8.5 window, footprint max r 32.2 < race ID 34 |
| 3a | Deck pass, drop below the deck | OBSTRUCTED (Ø5) / marginal (Ø3.6) | pan pedestal refills the pass to a 4.0 mm window |
| 4 | Seat void jog → platform slot → channel base | CLEAR | 3.93 mm |
| 5 | Neck channel, z 66 → 125 | CLEAR | 4.00 mm (16.5 × 8.5 window at z 95) |
| 6 | Channel exit at the column top | OBSTRUCTED | 1.0 mm escape gap (riser + root block + arm + worm box it in) |
| 7 | Tilt drape, column top → head slot → Pi | CLEAR (once 6 is fixed) | 0.32 mm chord graze at −30° (wire bows; all worm parts ≥ 10 mm) |
| 8 | Head-side, slot → Pi USB-C vs worm/wheel/axle | CLEAR | worm 14.2, wheel 19.4, axle 23.5, stack 12.3 mm |
Two geometry fixes needed (both in src/build.py, exact cuts below). Defect 6 is
pre-existing (identical relative geometry before the drop, the whole
column-top/riser/worm cluster moved together), not caused by the drop. Defect 3a
is a build-order bug.
1. USB port → cavity, CLEAR
The 14×12×8 port cut at (0, −78, 19) opens the rear wall (containment: wall material absent at (0,−77.4,19)). Min clearance of a Ø5 conduit from the port into the cavity: 4.00 mm.
2. Service loop, CLEAR, but it lives in the TOP band of the cavity
Full-ring annulus scan around the pan axis (5° steps, 2 mm radial steps):
- z 38 and z 43: annulus r 40–52 and r 44–56 fully free over 360°.
- z ≤ 30: blocked at azimuth 190–215° (TT drive motors) and 330–345° (ULN standoff corner). Worst free radial window at z 26 is still 15 mm, but not at a loopable radius all-round.
- r < 40 blocked at all heights by the pan pedestal (48×48 at (−7.9, 0), top z 44.25, only 1.75 mm under the deck).
So the 2-turn Ø3.6 loop coils at r ≈ 44–52, z ≈ 37–45.5 (radial window 12–16 mm, vertical 8 mm, two turns side-by-side need 7.2 mm: fits). 2 turns at r 48 ≈ 600 mm of wire; pan ±90° winds/unwinds ~±150 mm (half a turn): the loop absorbs it with a full turn of margin. Free annular volume r 20–64, z 14–45: ~330 cm³ (89 % of the annulus is free). Pan motor / pedestal / ULN don't crowd the loop band; they only push it up against the deck underside.
3. Deck pass, the deck cut is fine, the PEDESTAL blocks the drop (defect A)
Through the deck floor (z 46–51) the 16×8 obround at (12,−24) is fully open (probed
16.0 × 8.5 mm) and its footprint max radius is 32.2 < race-ring ID 34 (nearest
approach to pan_race: 7.13 mm). Aligned with the platform slot at pan=0.
Below z 44.25 the pan-motor pedestal refills most of the pass footprint. In
build_base() the cable pass cbl is subtracted (~line 1294–1298) BEFORE the
pedestal is unioned (~line 1300–1306), so the pedestal (x −31.9..16.1, y ±24,
z 12..44.25) fills the pass back in over y > −24. Probed free window under the deck
at x 8–16: y −28..−24 ≈ 4.0–4.3 mm (plus full 8 mm only for the x 16.1..19.9
sliver). A vertical Ø5 conduit at (12,−24) has 0.29 mm to the pedestal corner: FAIL.
The Ø3.6 pair squeezes through the 4 mm window: marginal.
Fix: move the cbl subtraction to AFTER the pedestal union (or subtract cbl a
second time there). cbl already spans z 29–51, so this bores a clean 16×8 shaft
through the pedestal corner; below z 44 the shaft's −y side is open to the cavity
(pedestal stops at y −24), so the wire swings straight back into the loop band.
No other part is affected (overlap zone is pedestal corner x 4..16, y −24..−19 only).
4–5. Seat-void jog + platform slot + neck channel, CLEAR
Jog polyline (12,−24,50) → seat void → platform slot → (0,−26,68): min clearance 3.93 mm (platform), 7.1 mm (race ring), 10.8 mm (balls). Platform slot and neck channel are mutually aligned at pan=0 (both aim along (0,−26)→(12,−24)); slot and channel windows probed at 16 × 8.5 mm. Channel clear at 4.0 mm from z 66 to the column top at z 125. At pan≠0 the platform slot + channel rotate together (same pan group) so they never misalign; the fixed deck pass ↔ rotating slot offset is taken by the wire winding in the 7.4 mm seat-void disc (z 51–58.4, r<34) + the loop below.
Minor: the 270° neck-bolt M3 head counterbore (Ø6.5 at (0,−33)) breaks into the
platform cable slot edge by 0.25 mm (slot edge y −30.0 vs cbore edge y −29.75).
Confirmed by containment probe. The screw head loses a sliver of its seat,
functional, but if touched anyway: bump that bolt circle rad 16.0 → 16.5 in BOTH
build_neck_clevis() and build_pan_platform() (they must stay in sync).
6. Channel exit at the column top, OBSTRUCTED (defect B, the real find)
The channel emerges at the column top (z 125) into a chimney that is boxed in on every side (all containment-verified on the built mesh):
- front: the worm-cradle riser (x ±9, y −25..−20, z 125–141.5): the channel cut hollows its back 3 mm, but its front 2 mm wall (y −22..−20) survives full-width;
- rear: the cheek-root block (y −40..−26, z 125–135): bored only for |x| ≤ ~8;
- sides: at z 129.25–135.25 the cradle arm (x ±9) closes the flanks; on +x the gusset (x 14..24, z 122–144) blocks the right bay;
- top: the worm thread envelope, clearance at (0,−26,z) falls to 2.46 mm at z 134 and goes negative at z 138 (worm bottom ≈ 135.6).
Measured escape gaps: left, between root face (y −26) and riser back (y −25): 1.0 mm (z 125–129.25 only). Forward, in front of the riser: y −20..−18.2 = 1.8 mm, and the −30° head sweep (display stack rear digs to y −18.2 over z 123–135 post-drop) closes onto it. Both fail even the Ø3.6 bundle. The wire cannot legally leave the channel.
Fix: a side exit window in the column's top-left corner, subtracted from the
neck after the main union (next to the existing chan cut in build_neck_clevis(),
~line 1014): box(12, 8, 10) centered (−12, −26, 122) → spans x −18..−6,
y −30..−22, z 117..127. Probes confirm the zone is pure column/riser-bottom-corner
material; nearest other parts are motor_tilt at 10.9 mm and the platform 55 mm below;
it leaves the x −24..−18 column corner and a 2.5 mm web to the chin-notch face
(y −19.5). The cable then exits at x ≈ −12 below the riser into the left bay
(x −9..−24, y −26..−20), which is open upward to z ~147 (probed window at
(−13,−22,127): 20 mm free up, 20 mm free left) and stays 1.8 mm clear of the −30°
head sweep (bay is behind y −20). Mirror on +x is NOT available (gusset).
7. Tilt drape, CLEAR through the slot at all angles (given the fix above)
Anchors: pan-frame exit X1 = (−13,−22,127) (top of the left bay); head-frame entry H = Pi USB-C at local (−30,−7.5,126) (bottom edge of the stack; the ETH edge faces +X so power is at the −X end, the plug sits 1 mm inboard of the head slot's x ±31 side wall).
| TILT | H (world) | straight gap X1→H | chord min-clearance |
|---|---|---|---|
| −30° | (−30, −22.4, 124.4) | 17.2 mm | 0.32 mm (neck root, wire bows forward) |
| −15° | (−30, −14.9, 124.2) | 18.7 mm | 0.95 mm |
| 0° | (−30, −7.5, 126.0) | 22.4 mm | 1.82 mm |
| +15° | (−30, −0.9, 129.6) | 27.3 mm | 2.00 mm |
| +30° | (−30, +4.6, 134.9) | 32.5 mm | 2.00 mm |
All worm-drive parts stay ≥ 10 mm from the chord at every angle; head shell ≥ 14.3; the bottom-rear slot faces the exit across the whole range (exit stub vs head shell: 23.9 mm at −30°, 18.3 mm at +30°, no pinch). Total wire, column exit → plug: 31 mm (−30°) → 46.5 mm (+30°), i.e. ~15.5 mm of drape travel. Install ≈ 60 mm of free lead (max 46.5 + bow + strain relief at the plug); the slack bows down-forward into the open head cavity at −30° and straightens at +30°.
8. Head-side path vs the worm/wheel/axle zone, CLEAR
Neutral-pose polyline (−13,−22,127) → (−13,−12,127) → (−20,−8.5,126.5) → (−30,−7.5,126), i.e. hug z ≈ 126–127, cross under the −X cheek at y ≈ −8.5: worm 14.2 mm, wheel 19.4, axle 23.5, motor_tilt 14.5, neck 2.0, head walls ≥ 18.7, screen/Pi stack 12.3 mm (unsigned, reference mesh isn't watertight). The whole mechanism cluster sits at z ≥ 135.6; the cable never needs to climb above z ≈ 128 until it is forward of the worm, so tilt motion doesn't sweep the drive through the wire. At −30° the swept stack bottom passes 10.5 mm above the drape chord.
Action list
build_base(): subtract the deck cable passcblAFTER the pedestal union (defect A, restores the full 16×8 drop; today's real window is 4.0 mm).build_neck_clevis(): add the side exit windowbox(12,8,10)at (−12,−26,122) (defect B, today the channel exit is sealed to a 1.0 mm gap).- Optional: re-clock/bump the 270° neck-bolt circle 16.0 → 16.5 to un-graze the platform cable slot from the M3 head counterbore (0.25 mm overlap).
- BOM/assembly note: route the power pair with ~60 mm free lead between the column exit and the Pi plug; coil 2 turns (~600 mm) at r ≈ 48, z ≈ 38–45 in the cavity; keep the pan software limit at ±90°.