3a. Tabs — nest-time hold & break-off

kerfmaster pin list · Big capabilities

rulingmixedA decision by Aristide or Jordan. True because it was decided; it can be superseded, but it cannot be stale.

Content last changed 2026-08-13 — computed from the item itself, not typed.

The problem. A part cut fully free of the sheet can drop, tip, or shift into the machine before the job finishes — wrecking the part, the cut, or the head. Tabs are small connections deliberately left between the part and the surrounding sheet (the skeleton) so the part stays put while cutting, and are broken off afterwards in the manufacturing process.

⚑ A tab is defined by the fact that it breaks. If a connection is meant to stay in the finished part, it is a bridge (item 2b), not a tab.

Where it sits in the pipeline

Tabs are a nest-stage feature, not a part-stage one: they connect the part to the sheet, so they cannot be placed until the part has a position on the sheet. Order: … → bridges (per part) → leads/cut-order → NEST → tab placement → post.

• Contrast with bridges, which are per-part and baked in before nesting.
• Same post-time mechanism as a bridge: a gap left in the cut path (beam skips the span). The difference is intent and size — a tab is sized to fail easily.
• A lead-in / pierce must never land inside a tab span (same rule as bridges).

Sizing

Jordan: tab values vary — different thickness/width values are used at nest time. So this is a table lookup, not a formula: material × thickness → tab width, with a floor (what the beam can cleanly leave) and a ceiling (before it stops breaking easily).

Note the sizing goal is the inverse of a bridge: a bridge is sized to hold, a tab is sized to hold just enough and then snap cleanly by hand or in handling.

Settled so far

• Part → skeleton always. Tabs attach the part to the surrounding sheet. Part→part tabs are out of scope — two parts will almost never share a cut line (Jordan). This removes shared-edge handling from the nester entirely.
• The tab width table is coming from the shop — Jordan and Aristide are preparing it as homework, then we refine it together. Not a blocker for wiring the stage in.
• Tab count rules will be derived from real examples the shop feeds us, rather than guessed from a formula up front.

⚑ Still open

⚑ (1) The tab width table itself — material × thickness → width, plus the floor (thinnest the beam can cleanly leave) and ceiling (widest that still snaps easily). (2) Tab count + spacing, to be inferred from the shop’s examples. (3) Placement — are tabs kept off finished/visible edges, given the break leaves a witness mark? (4) Does the leftover nub get cleaned up after break-off (see below)?

“Dressing” = cleaning up the small leftover nub/witness mark after the tab is snapped — grinding, filing, sanding or deburring it flush. It matters to us because if the mark is dressed afterwards we can afford slightly larger tabs in more visible places; if it is not, tabs must hide on edges that don’t show in the finished part.

src — Jordan, 2026-07-28: terminology correction (tab = breaks off, bridge = stays) and answers round 1 (part→skeleton only; width table and count rules coming from the shop).

When a part needs a tab at all (Jordan, 2026-08-10)

Tipping is a function of material thickness vs kerf. A part can only tip by cocking into its own kerf slot; on thick material the kerf (~0.007") is tiny against the thickness, so e.g. 1" plate can never tip and needs no tab. Thin sheet can, and does.

• Default per material × thickness lives in the material table — whether parts get tabbed, and (when the shop numbers arrive) width and count.
• Operator override is per part, in the interface — interface intent, not built; the default must be changeable part by part, per the standing every-default-overridable rule.

2026-08-12 — first shop number, measured from 33199.NC: their MetaCam post for our -61 nest tabs ALL 48 part outers with a flat 0.030 in gap on 14 ga (MSO7,0.080), every span on a flat, none at a corner (48/48 audited). Because parts are held, they cut part-by-part — 826 in of rapid vs our 1134, and zero rapids over loose pieces vs our 42 in the outer endgame. Wiring part tabs into our post awaits Jordan's explicit go (asked same day).

2026-08-13 — GO given by Aristide (partner rulings bind equally): wire part-outer tabs into our post. Placement follows the standing start-point rule (on a flat, centred, lead-aware, never at a sharp point); pierce/lead never inside the tab span. Width table so far has ONE measured row: 0.030 in on 14 ga (MSO7,0.080), on a flat — other material × thickness rows come from shop numbers or measured from their NC, never guessed. Tabbed parts also unlock part-by-part cut order (parts are held, so rapids over them are safe until severed).

2026-08-13 — WIRED IN (record). emit.write_gcode_macrob(part_tab=…) leaves the uncut span on every depth-0 boundary through the same start-rule/lead machinery as slug tabs (on a flat, centred, lead-aware); width resolves through tables.part_tab() from materials.json (MSO7,0.080: 0.030, provenance measured); a missing row raises. post_nest passes it on every nest post. Tabbed outers join the hazard set: the order scorer and the endgame now count rapids over held parts honestly — TL-1016 re-post shows 197/197 outers held at 0.030, 0 spans unfit, audits clean (3 pre-existing narrow-feature pierce listings unchanged). Crossings ROSE on the two dense nests (3→18, 0→43) because flights over severed parts were previously loose AND uncounted (the retired known-limit note); they are now held and counted. Next lever, intent: part-by-part cut order as in their 33199.NC, which runs zero.