Tip Profile Study

A retracted claim, kept here on purpose

An earlier version of this page reported that v4's mounting-hole spacing was wrong — that the rigid mount measures 12.000 mm across every slice, that the blade's inherited holes match it, and that v4's 10.0 mm therefore had to be a mistake. The file measurements were real. The conclusion was wrong, and it has been fully reverted.

The operator measured the physical hardware with a caliper: the real screw spacing is 10.0 mm. The 12.000 mm in the STLs is a known defect, ruled on 2026-08-19 — the uploaded CAD does not match the machine. All the files that "agree" on 12.000 descend from that same bad upload, so their agreement carries no weight.

The lesson is the one already written on the FK bridge in this project: a spec or an upstream file can predate the measurement, and the measurement wins. Now extended — three files corroborating each other is not evidence when they share an ancestor. The shipped STL was restored byte-for-byte from git, and generate_v4_finish.scad now carries a warning against making this change again.

Still unresolved: the lateral shift, and therefore closure

2B also asks for the whole hole pattern to move laterally inward, because the holes sit too far toward the outer edge of the TPU foot and that forces each jaw inward on the mount. The recorded consequence: "through part of the closing motion, the two long fingers can approach a geometry where they would overlap/cross rather than closing cleanly."

That offset is HOLE_DX and it is still 0 — unmeasured. So whether two tips meet cleanly at full closure is not guaranteed, and no tip profile can be judged on a pair that is not aligned to begin with. The original blade STL carries no mounting holes at all, so this one cannot be derived from files.

01 — The current profile

Why it reads as a trapezoid

Protrusion measured along the inward normal, d = distance back from the tip. Three constraints stacked up to produce the plateau.

  d (mm from tip)   inward (mm)   profile
      0 – 9            0.00        dead zone — keeps the tip sharp (1.13 mm)
      9 – 11        0 → 3.50        ramp in
     11 – 19           3.50        ████████ 8 mm flat — where two jaws mate
     19 – 21        3.50 → 0        ramp out
     21 +              0.00

Constraint	Effect on the shape
The tip must stay sharp	An earlier build carried the lip at full 3.5 mm depth all the way to the point. That added 3.5 mm of thickness at the tip and made it blunt — rejected on inspection against the real red gripper. The fix was a 9 mm dead zone.
Two jaws mate on a flat	`LIP_FLAT = 8` exists so the pair closes face-to-face rather than on a line.
A 90° step will not print	Both sides get a 2 mm ramp; a vertical step is an overhang in TPU.

The distinction that got lost: the earlier failure came from carrying the lip to the tip at full depth, not from reaching the tip at all. A protrusion that starts at zero at the point and grows rootward keeps the tip just as sharp — which is what the alternatives below do.

02 — Three profiles, same mesh pipeline

Rendered in profile, tip at the top

Each is a real generated mesh, not a sketch. A is the shipping v4; B and C change only the tooth() function.

Trapezoid — current v4 shipping

Dead zone 0–9 mm, ramp 9–11, 8 mm flat at 3.5 mm, ramp out 19–21.

The step between the point and the plateau is what reads as blunt. The 8 mm flat is the only face two jaws can mate on.

Barb triangle proposal

Rises from 0 at the point to 3.5 mm at d=15, then falls back over 2 mm.

Continuous — no step anywhere. But the return face sits at 60.3° to the rail, not square. No flat mating face.

Wedge proposal

Rises from 0 at the point to 3.5 mm at d=21 and stays.

The gentlest continuous taper and the largest bearing area — but the protrusion never returns, so it carries on rootward. No flat mating face.

03 — The trade-off

A continuous taper costs contact area

The 8 mm flat is what lets the two jaws close face-to-face. B and C replace it with a slope, so a closed pair meets along a line rather than a face — and the original field-test complaint about v3 was precisely that contact area was too narrow and the cup slipped out. C keeps more area than B by never ramping back down.

Variant	Tip sharp	Continuous	Flat mating face	Risk
A — trapezoid (current)	yes	no	8 mm	reads blunt; step between point and lip
B — barb triangle	yes	yes	none	line contact; least grip area
C — wedge	yes	yes	none	line contact, but the widest bearing slope
D — barb, 90° hook face	yes	yes	none	the operator's pick; hook face square to the rail

04 — D: the hook face square to the rail

What the operator actually asked for

B was close but its return face sits at 60.3° to the inner rail. A hook that bites wants that face perpendicular. D rises from zero at the point to 3.5 mm over 15 mm (13.1° from the rail), then stops dead — the hook face measures 89.7°, square within the 0.02 mm station spacing used to build it.

✓

The "a 90° step won't print" objection does not apply here

That comment has been in the source since the trapezoid was designed. But BLADE-README.md §摆放 specifies the blade prints flat — the 10 mm face on the bed, layer lines perpendicular to the bending direction, because Fin-Ray works by rib buckling and layer adhesion is the weak axis. Laid that way the entire lip profile lies in the bed plane, so a 90° step is a vertical wall with zero overhang. The 2 mm return ramp was an unnecessary concession.

Mesh checks pass: shells 1, watertight, manifold, no degenerate faces, signed volume 25,696.24 mm³.

D — barb, 90° hook operator's pick

0 at the point → 3.5 mm at d=15 (13.1° from the rail), then cut square.

Hook face measures 89.7° to the inner rail. The taper to the point keeps the tip sharp; the square face is what actually catches the cup. Mesh passes: shells 1, watertight, volume 25,696.24 mm³.

What has to be measured before this can be decided

Four numbers, all with a caliper on the real parts. Nothing here can be taken from a file.

Original red jaws, installed, fully closed. The gap between the two grip faces at the tip, and at the point where they first touch. If they touch flat, say roughly how long that contact is.
Current long jaws, installed, fully closed. The same two numbers — and whether the two fingers cross rather than meet.
Black rigid mount. Screw-hole centre-to-centre spacing (should confirm 10.0 mm), and the distance from the screw pair's centre to the mount's inner face.
Two blades held inner-face to inner-face. The outside-to-outside span across the two feet — compare against the same measurement on the original pair. That difference back-solves the hole shift directly.

Give me any subset and I will compute HOLE_DX, put it in generate_v4_finish.scad, rebuild, and draw the closed top view for the original and the new pair side by side.