SPACE ENGINEERS AUTO SHIP-BUILDER — DILEMMA SUMMARY Saved 2026-06-28 for pickup in a fresh session. Companion diagram: https://infographics.horoszow.ski/se-autobuilder/se-autobuilder.html (7 panels) ================================================================== THE GOAL ================================================================== Build an automated printer that welds a projected blueprint layer-by-layer. - First test piece: solid light-armor block, 7 x 10 x 4 (280 blocks, 4 layers). - Real target: a SMALL-GRID atmospheric mining ship. - Base is Peter's MAIN base, STATIC grid (no sag/Clang worries from gravity). ================================================================== CURRENT RIG (confirmed from screenshots SNAG-0006/0007) ================================================================== - A row of welders on a horizontal beam, pointing UP at the projection. - Beam is bolted to the top of ONE vertical (Z) piston column. That is the ONLY powered axis. NO horizontal/retract piston. NO sweep axis. - Feed (components) piped in via conveyor; seed/anchor is a CONNECTOR. - Peter's chosen motion model: WELDERS MOVE, seed/build stays fixed. - Peter has now decided: he is NOT lifting/indexing at all — he only wants to RETRACT the welder. (This is the crux of the open question — see below.) ================================================================== THE CORE PROBLEM (the jam) ================================================================== Welders weld via a SPHERE (~radius covering the active layer), centered on the welder BLOCK — not a beam from the tip. A powered welder welds EVERY supported projected block in range, continuously. Consequences: - For a SOLID build, a stationary powered welder fills cells toward ITSELF and walls itself in / collides when it tries to move = STUCK. - Moving the welder body straight UP through a solid drives it into cells the build just filled = STUCK. (This is what Peter has been hitting.) ================================================================== PETER'S KEY INSIGHT (and where we landed) ================================================================== Peter correctly argued: if the welder is IN the build plane and powered, you CANNOT win by out-pacing weld-speed — it's an unwinnable timing race. Correct. RESOLUTION reached (to be implemented next session). The race is removed by TWO things together, NOT by fast timing: 1. DEPTH STANDOFF — the welder must never occupy a build cell. It sits OFFSET in the build-growth direction, reaching INTO the active layer with its sphere while its body stays in empty space. No blueprint blocks exist in the gap, so nothing welds there; the welder can never be enclosed. - Flat/shell builds: depth standoff ALONE solves it (sweep the face, no index). - SOLID builds: each layer grows toward the welder, so you ALSO need step 2. 2. WELDERS OFF DURING THE INDEX MOVE. Cycle for a solid: weld layer (ON) -> welders OFF -> open the gap (move build away OR move welder back one layer) -> welders ON -> next layer. With welders OFF while repositioning, nothing builds during the move => NO RACE. Timing becomes loose: just ensure the layer finished (generous timer, or read projector "remaining blocks"), not split-second. The bug in the current rig is most likely: welders left ON continuously + no standoff/retract, so the welder keeps building into the cell it must move through. ================================================================== OPEN QUESTION TO RESOLVE NEXT SESSION ================================================================== Peter said "I'm not lifting ever, I'm only retracting the welder." Need to reconcile this with a SOLID build: - Retract-only (pull welder back in depth) works PER LAYER but you still must index ONE layer between passes (either retract further each layer, or the build must advance). Clarify exactly which axis indexes for his solid test. - Decide between the two robust designs (see below). ================================================================== TWO ROBUST DESIGNS ON THE TABLE ================================================================== OPTION A — keep movable welders, ADD a horizontal RETRACT piston: Insert a horizontal piston between the Z-column top and the welder beam. Cycle: weld -> retract welders OUT of footprint -> index one layer -> extend back to standoff -> weld. Bodies travel only through empty air. Smallest change to what he already built. OPTION B — fixed welders, move the BUILD (sturdier, scales to any height): Bolt the welder beam to the fixed frame; put the connector/seed on the Z-piston so the BUILD rises away from stationary welders into empty space. No burial possible, no retract axis needed. More re-plumbing. BEST FOR TALL/COMPLEX (the miner): perimeter welder RING — welders around the hull cross-section, bodies OUTSIDE the profile, build extrudes THROUGH the ring plane. Welders can never be buried regardless of height. (Panel 7.) ================================================================== CONTROL WIRING (already designed — panel 5) ================================================================== Vanilla, no Programmable Block: - T_WELD timer (delay = weld time): WELDERS Off; LIFT velocity +0.5; start T_STEP. - T_STEP timer (delay = lift time): LIFT velocity 0; WELDERS On; start T_WELD. - EC_DONE Event Controller: watch piston position; at final layer -> stop T_STEP, start T_FINISH. - T_FINISH: one last weld window, then WELDERS Off + alert. Crucial: WELDERS turn OFF during the reposition phase (this is the anti-jam). For Option A, the loop must add the RETRACT + EXTEND piston steps (peck cycle). GRID PITCH: large grid = 2.5 m/layer; SMALL grid = 0.5 m/layer (redo lift-time math for the miner: e.g. 0.5 m/s velocity -> 1.0 s per small-grid layer). ================================================================== SMALL-GRID SEED (for the atmo miner — panel 6) ================================================================== - Can't build small-grid directly onto large grid. Bridge with a large ADVANCED ROTOR set to a SMALL HEAD (vanilla: detach large head -> "Add Small Head" button). Build the small seed (projector + connector, optional battery) on that head. Lock the rotor rigid. - Power CROSSES the rotor (projector stays alive, often no battery needed). Conveyor/items do NOT cross a rotor — feed components via the seed's connector. - Print onto a CONNECTOR (not landing gear): clean undock + the finished miner gets a working dock port for free. (Merge block also works; connector better for a ship you fly/dock repeatedly.) - Projector should be on the FIXED frame so the projection stays nailed at the weld plane; only the seed/build should ride a moving axis. ================================================================== SIDE NOTES ================================================================== - Airtightness: Air Vent status. "Pressurized" + green = sealed & filled. "No leaks detected" but NOT pressurized = sealed room but no O2 to fill it -> check (a) Depressurize toggle is OFF, (b) an O2 source (O2/H2 generator with ice, or filled O2 tank) is on the connected conveyor net, (c) vent's flat back face is mated to the conveyor network. Don't need a generator adjacent, just reachable on the network. - DELIVERY/VIEWING FIX (infra): SMB share (H:) caches the directory listing on Windows (SMB2 dir leases) so new/edited files don't appear without a disconnect/reconnect, and Chromium refuses file:// on mapped network drives (ERR_FAILED) + copy threw 0x80070057 on a stale inode. SOLVED by serving the ~/hermes-inbox/infographics dir over HTTP via a new nginx:alpine container ("infographics") behind Traefik at https://infographics.horoszow.ski (LE cert issued, HTTP 200 verified). snag (Windows->share write) is UNTOUCHED and still works; only the infographics subdir is served, read-only. View diagrams at the URL and just F5 — no more reconnect dance.