Anchor-bulb pre-tensioning DOE: 15-specimen test plate (air gap × joint size, A3 countersunk, horizontal cable)

[Copilot]

Pre-tensioning the anchor-bulb requires a PLA↔TPU bore bond weak enough to pull through under finite force, and an air gap small enough that the TPU still prints as a clean cylinder when bridging the bore horizontally against gravity. This adds a single H2D plate that sweeps air gap × joint size for the A3 countersunk anchor head so we can pull-test each cell and pick a winner in one print job.

cad/anchor-bulb-tensioning-array/

• _common.scad — A3 countersunk joint (90° conical TPU head mating a 90° countersink cut into the +Y face of the PLA node, flush + self-centring + ~8.8 mm² conical bearing wall) + 16×26×4 mm pull-test tab, parameterized on (node_d, gap_r). Per-material modules (*_pla / *_tpu) and a colour-preserving X=0 section module.
• 15 specimen files — H-S{0,1,2}G{0..4}.scad, one per DOE cell with embossed ID on the tab:
  • Joint size (rows): node Ø 7.5 / 9.5 / 12.0 mm (S0 / S1 / S2). S1 = PR Explore and validate joint design for PETG/PLA + TPU prints (5 candidate joints + OpenSCAD CAD + anchor-upset shape variants + Design F captive-core + Edison Phase-1/2/3/5 + Phase-4 submitted + lander-context reco + TPU-inside vs TPU-outside comparison... #39 Phase-3 default; S2 = matches dovetail node OD; S0 = smallest with ≥2 PLA wall perimeters at the equator. Bore length tracks node Ø, so this axis tests whether a longer horizontal bridge tolerates a different gap.
  • Air gap (columns): radial clearance 0.1 / 0.2 / 0.3 / 0.4 / 0.6 mm (G0 … G4); bore Ø = 2.4 mm cable + 2·gap.
• tensioning_array.scad — 3×5 grid, 22×60 mm pitch, ~110×130 mm footprint.
• render.sh + renders/ — per-specimen iso PNG + STL, 3 X=0 cutaways (one per node size at mid gap = 0.3 mm), full-plate iso + STL, 2 contact-sheet montages.
• README.md — DOE table, "what changed from the first revision" change log, IDEX (two-extruder) H2D / AMS-Pro-as-filament-store recipe, horizontal pull-test protocol with success criterion F_pre ∈ 5–25 N, and a forward note on the manual-threading alternative.

// _common.scad — two knobs: joint size and air gap
specimen_A3(id     = "H-S1G2",
            node_d = 9.5,   // axis 1: PLA node Ø (= bore length)
            gap_r  = 0.3);  // axis 2: radial air gap (mm)

Layout choices

• A3 countersunk head (not A1 frustum): flush, self-centring, ~2.6× the bearing-wall area at the same OD.
• Horizontal cable orientation — cable axis lies in the build plate (+Y) with the strut going up (+Z). The TPU has to bridge the bore horizontally as it extrudes, with gravity pulling it down into the air gap — i.e. intentional spaghetti / stringing across a short span. This is the real failure mode (the prior vertical-cable layout hid it).
• Single interface-treatment axis (air gap) — the H2D is IDEX (two extruders), not three nozzles, and the AMS Pro 2 is a 4-slot filament store on the L extruder, so a sacrificial-PVA 3rd material is not available. Pause+lubricate is a manual workflow that doesn't survive a multi-specimen plate cleanly. Both were dropped from the first revision.
• Self-contained directory; does not include from PR Explore and validate joint design for PETG/PLA + TPU prints (5 candidate joints + OpenSCAD CAD + anchor-upset shape variants + Design F captive-core + Edison Phase-1/2/3/5 + Phase-4 submitted + lander-context reco + TPU-inside vs TPU-outside comparison... #39's cad/joint-design/A_variants/, so mergeable independently of Explore and validate joint design for PETG/PLA + TPU prints (5 candidate joints + OpenSCAD CAD + anchor-upset shape variants + Design F captive-core + Edison Phase-1/2/3/5 + Phase-4 submitted + lander-context reco + TPU-inside vs TPU-outside comparison... #39.
• The manual-elastic-threading alternative (no co-print needed) is captured in the README's "Notes / next steps" as a complementary path — this array remains useful for the co-printed-tendon direction (Add strut-curvature design-axis exploration synthesis #54), where we still need to know the limiting air gap.

Renders

X=0 cross-sections (section_montage.png) show the bore + countersink + conical head at each node size. OpenSCAD 2021.01 preview-mode shades cut interiors dark — boundary edges still convey PLA / TPU layout.

[sgbaird-alt]

Also as a personal aside, just realizing after reading https://github.com/vertical-cloud-lab/tensegrity-optimization/blob/copilot/suggest-tensegrity-kits/reviews/tensegrity-build-kits.md#22-tendons-tension-members, if we're going to try to pretension manually by pulling/clipping something, we may as well just leave openings and manually thread in an elastic material in the pre-specified locations, assuming we could get it to compress and thread through well enough (seems likely with it being an elastic material). We would then print supports for otherwise "floating" or unsupported objects (due to removing the cables from the print). This also means we wouldn't necessarily even need to use a multi-material printer. We can just use a regular printer. Obviously more manual work involved, but frankly not that much. In the future, as we move towards #54, this workaround may not be as applicable, but even then I'm finding it hard to find clear justification for why an ad-hoc installation of cables would be a bad idea, unless we get to the point where we can test automatically without human intervention (i.e., automatic removal from build plate).

[sgbaird-yolo]

@copilot noting that when it's vertically oriented, of course it's going to be fine if we leave a large enough gap. What we're worried about it when it's oriented horizontally, so it's essentially intentional spaghetti / stringing (but with a very low distance for it to drop) in the case of an air gap. Also, we don't have any PVA (and you're mistaken, there's no 3rd nozzle in the H2D, just that the Bambu Lab AMS Pro 2 has 4 slots for different types of filaments). So your current set of designs doesn't help us. Focus on the air gap for different air gap sizes and different joint sizes. Also, you made a mistake using the ball shape. What I actually want to use is "A3 countersunk — 90° cone mating a countersink in the +X face → flush, self-centring, ~8.8 mm² conical wall" (originally I specified A1, but I think this A3 from #39 (comment) makes more sense perhaps)

[Copilot]

@copilot noting that when it's vertically oriented, of course it's going to be fine if we leave a large enough gap. What we're worried about it when it's oriented horizontally, so it's...

Redesigned per these points in d7be0e5:

• A1 frustum → A3 countersunk (90° cone + matching countersink in the +face).
• Horizontal cable orientation — cable axis lies in the build plate (+Y) with the strut going up (+Z), so the TPU has to bridge the bore horizontally and gravity pulls it into the gap. That's the failure mode the DOE now exposes.
• Dropped PVA sleeve and pause+lube axes; corrected the H2D description (IDEX dual-extruder, no 3rd nozzle; AMS Pro 2 is a 4-slot filament store on the L extruder).
• New DOE: air gap × joint size — 15 specimens, 3 node sizes (7.5 / 9.5 / 12.0 mm) × 5 gaps (0.1 / 0.2 / 0.3 / 0.4 / 0.6 mm). Bore length tracks node Ø, so the size axis tests whether a longer horizontal bridge tolerates a different gap.

Full plate render:

3 × 5 contact sheet (rows = node size, columns = gap):

Also as a personal aside, just realizing after reading https://github.com/vertical-cloud-lab/tensegrity-optimization/blob/copilot/suggest-tensegrity-kits/reviews/tensegrity-build-kits.md#22-tendons-te...

Captured this as a complementary path in the README's "Notes / next steps" in d7be0e5 — the manual-threading workflow doesn't need an anchor bulb or multi-material printing at all, so it's mergeable independently of the array. The DOE in this PR stays useful for the co-printed-tendon direction (#54), where we still need to know the limiting air gap before the TPU stops bridging the bore cleanly.