← all reports
everything still unmeasured one session with a caliper 2026-08-21

The Five Measurements

All five are read in one state — gripper fully closed, blades installed — and together they fix everything still open on the gripper: how far apart the blades sit, which way they converge, and whether the two protrusion faces can meet as faces rather than as a line.

Why five and not two

Two distances only fix a translation. The two mounting patterns can also be rotated relative to each other, and the URDF says they are: at full closure the two jaw axes sit 10.79° apart. Two distances plus one angle pin the relative pose completely — and the inner-bolt offset as components rather than a straight-line distance, because when two lines are not parallel a single distance between them is ambiguous.

x = 0 · centreline x y left blade right blade θ L_out L_in R_in R_out P1 P4 TL TR T C — Δx and Δy between the two INNER bolts B — outer bolt to outer bolt A — foot outer corner to foot outer corner
Six named points and five readings. Blue belongs to the blades, orange to the mounting bolts. The two dashed extensions meet where the bolt lines cross — the angle between them is θ. The drawing exaggerates the convergence so the geometry is visible; the real angle is around 11°.

01 — The six points

PointWhere exactly
P1Left blade — outermost corner of the mounting foot's bottom face
P4Right blade — outermost corner of the mounting foot's bottom face
L_outLeft base — centre of the bolt further from the centreline
L_inLeft base — centre of the bolt nearer the centreline
R_inRight base — centre of the bolt nearer the centreline
R_outRight base — centre of the bolt further from the centreline
TL / TRThe two blade tips

02 — The five readings

 ReadWhat it settles
AP1 → P4, horizontalWhere the feet sit — with B, gives the bolt-to-foot-edge offset
BL_out → R_out, horizontalWhere the bolts sit. (A − B) / 2 compares against the CAD's 5.004 mm
CL_in → R_in, as Δx and ΔyPins the two bolt patterns' relative position exactly, independent of A and B
TTL → TR, horizontalWith A, gives the convergence: atan((A − T) / 2 / 89.4)
θAngle where the two bolt lines cross, extendedThe mounting-plane angle. Also say which side they converge toward — tips or base

If a pair of points is not level, take the horizontal component — square across, not diagonally. 89.4 mm is the foot-bottom to tip distance from the CAD (z −33.985 to 55.44).

03 — What each one will be checked against

Every reading has a prediction attached, so a mistake shows up immediately

ReadingPredictedWhere the prediction comes from
(A − B) / 25.004 mmMeasured off Blade_v4_final.stl: outer hole to foot outer edge. A mismatch is HOLE_DX
Each bolt pair spacing10.0 mmThe operator's caliper reading on the real hardware — a cross-check that the right points are being used
θ≈ 10.79°From the URDF joint at full closure. This is the real test — if the measurement lands near it, the kinematics can be trusted for everything downstream
A − T— openNo prediction. This is the one that resolves whether the mating face needs flattening or leaving alone
!

θ is the one that decides a design question, not just a number

The blade's mating face currently sits 5.93° off the blade's own axis, following the inner rail. Each jaw sits 5.40° off the mirror plane. If those two tilts oppose each other they nearly cancel (−0.53°) and the face is already almost parallel — flattening it would make things worse. If they add, the real error is 11.33°, twice what flattening corrects.

A flattened variant is already built and passes every mesh check — mating face 0.036° off parallel, down from 5.93°, with the tip still 0.99 mm sharp. It carries a single switch, FLATTEN, that can be set to 0 to fall back or negated to reverse the correction. θ decides which.

Then it is one pass, not a conversation

With all five in hand: HOLE_DX is computed, the FLATTEN sign is settled, and the inward hook is already done. All three go into the source together, the three build steps re-run, check_mesh and check_ribs confirm it, and the closed top view gets redrawn with a correctly aligned pair — so the tip contact can be judged on an assembly that actually lines up.