Alcove rod socket

A parametric, two-part screw-together end socket for a 40 mm curtain or closet rod that spans a recess or alcove. A wall boss screws flat to each facing wall; the rod's ends plug into knurled collars that hand-thread onto the bosses. Curtains down for washing = unthread two collars, no tools.

v2 — ships the separable multi-object plate deliverable. v1 (#361) printed fused in the field (a single STL of the two parts slices as one welded body); v2 fixes the packaging, geometry unchanged. See Deliverable below and the field test log in NOTES.md.

What you get

Printed in pairs (one holder per wall, one pair per rod):

• boss — the wall plate: Ø58.8 × 18 mm disc with a countersunk M5 screw hole and an external printed thread (an optional second off-axis screw setting, screw_count=2, grows the flange for heavier installs — a hung pair over ~10 kg; see Assembly).
• collar — the rod socket: Ø54 × 40.6 mm knurled tube, internal printed thread below, Ø40.6 rod bore above.
• thread-coupon / bore-coupon — the "print this first" fit checks (see Print settings).

Deliverable — two objects, never one fused STL. boss and collar print as separate parts, and the one rule is: keep them as two distinct objects in the slicer. STL carries no object separation, so exporting the two into a single STL imports as one fused body and welds them together — v1's field-test failure (NOTES.md). Two ways to get the parts, both give the separation:

• Downloaded a Release: you get two files, boss and collar — import both and keep them as separate objects. (No v2 Release is tagged yet — releases are cut on a tag, not on merge — so until one exists, use the clone path below.) The Release carries the default-depth parts: for the push-in-deep / drop-in-shallow install, render the far-side collar shallow (engagement_depth=12) before install day.
• Cloned the repo: ./scripts/plate.sh alcove-rod-socket bundles the two into one multi-object 3MF (build/alcove-rod-socket-plate.3mf) — the same two objects in one file. (The 3MF is a local build artifact; it isn't attached to the GitHub Release yet — that's a tracked follow-up.)

Either way, saving bed space is fine — lay the two parts side by side and flat (the plate's --merge does exactly that). Don't stack them one on top of the other: two separate objects stacked vertically still print the upper one over open air — the ~10-layers-in spaghetti from v1.

A rod takes a pair — two holders, so print (or import) the set twice. For the push-in-deep / drop-in-shallow install (Assembly step 3, Parameters table) the two holders share the same boss; only the far holder's collar differs — render it shallow with engagement_depth=12. So a deep+shallow pair is two bosses + one default collar + one shallow collar. Neither shipped set — the Release nor the plate — carries that shallow collar yet; render it before install day with -D 'part="collar"' -D 'engagement_depth=12' (the part="collar" is essential — the file defaults to part="assembly", so without it you'd export the fused two-part preview as one STL, the very failure this fixes). Shipping the shallow collar as a gated part is the charter's #1 backlog item (B3 in PM.md).

Twice can be one print. Import both objects (or the plate), select both → right-click → Add Duplicate: four objects in one ~3 h 21 m run. For the deep+shallow pair, make the duplicate collar the shallow one — swap one of the two collars for a render with -D 'part="collar"' -D 'engagement_depth=12' — so the run is boss ×2, one default collar, one shallow. Two such plates fit the 256 × 256 mm P2S bed.

Print settings

• Material: PETG or ASA for a permanent, load-bearing install — PLA creeps under sustained load, and a hot south-facing window is the one-strike failure. Both want a retune, measured on the coupons: PETG prints the thread slightly fatter, so expect thread_tol +0.05–0.1 mm vs PLA; ASA shrinks the bore ~0.2–0.4 mm — about the whole slip — so print the bore coupon in ASA first and expect +0.1–0.2 mm on rod_clearance (a material-specific preset is charter backlog B4).
• Layer height: 0.2 mm, 0.4 mm nozzle.
• Perimeters: 3 — the 3.2 mm walls carry the load.
• Seam: Scarf (Bambu Studio) or Back — the default Aligned seam stacks a ridge up the printed threads and inside the rod bore, and that ridge can locally eat the 0.3 mm radial sliding fit.
• Infill: 20% gyroid.
• Supports: none needed anywhere, by design.
• Bed adhesion: brim the collar, per-object (Bambu Studio and most slicers brim per object — select just the collar). Its first layer is a thin ~2.2 mm annular rim carrying a 40.6 mm tube, the part that can lift; the boss doesn't need it — it prints flange-down on a full Ø58.8 disc, the best contact on the plate. Cheap insurance on the collar rim's own merits (the v1 field failure was packaging, not adhesion — see NOTES.md — so the plate is what fixes that; the brim is separate). A per-object collar brim can't reach the ~5.7 mm gap to the boss, so nothing bridges. (Brim the whole plate instead and the two brims may meet in one web; it should peel clean, but that is untested — predicted, not observed, and the proving print is what confirms it. Per-object avoids the question.)
• Orientation: print every part as rendered — boss flange-down, collar rod-mouth-down (the internal thread prints at the top of the collar, never on the first layer). Never flip the collar grip-band-down for a nicer face: the printed thread has to stay up top. If the collar's first-layer ring lifts, fix it with the brim above — don't flip the part.
• Print order: both coupons first, then tune (below), then a pair.

The gate scores three of the four gated parts 92/100 with one warning each — thin walls (2–5% of sampled surface under 0.8 mm on the boss, collar and thread coupon; the bore coupon is clean). That is the tessellated thread crests and the knurl ridges, not the 3.2 mm load path; it is by design and needs no slicer action. The coupon carries the same profile, so a coupon that prints clean says the parts will too.

Measure your pole with calipers where it will sit and set rod_d to that barrel reading — "40 mm" is sometimes the finial size, not the pole.

Parameters

The ones you are most likely to touch (all in alcove-rod-socket.scad, Customizer-grouped; override on the command line with -D 'rod_d=38.5' — and pair it with -D 'part="collar"', or you'll export the fused assembly preview instead of a printable part):

Sizes other than the default 40 mm rod are untested — the geometry scales, but the fits are only proven at 40 mm, and the coupons are the proof: print them in your material first and trust them over this page.

Assembly & use

1. Screw a boss to each facing wall, countersunk M5 head flush in the neck. Screw length: the boss is 18 mm through, so M5 × 40 mm into a rated drywall anchor (the anchor card governs the exact pairing), or M5 × 55 mm into a stud behind ~13 mm board (18 + 13 leaves ≥ 20 mm in the stud). One M5 in a rated anchor holds a pair to roughly 10 kg of hung weight — heavier curtains than that want a stud or screw_count=2 (derivation in NOTES).
2. Slide the collars over the rod before hanging it — one at each end.
3. Thread each collar onto its boss until the rim seats on the plate. For a rigid rod between two fixed walls: set one holder's engagement_depth deep and the other shallow (e.g. 12 mm) — push into the deep side, drop the shallow end in, tighten both collars. Cut the rod ≈ 18–20 mm short of the mouth-to-mouth span (28 mm deep − 12 mm shallow + each mouth's lead-in).
4. To wash the curtains: unthread the collars (~1¼ turns each — 2-start thread) and lift the rod out — holding the flange still as you unthread, because unscrewing friction can walk the single M5 out of the wall over enough wash days. No tools.

If a fit is off, don't resize the parts — reprint the coupon and move the tolerance (thread_tol, rod_clearance) in 0.05–0.1 mm steps; see "Print this first" in NOTES.md, including a one-print tolerance sweep (./scripts/render.sh alcove-rod-socket --sweep thread_tol=0.2:0.4:0.05).