Here’s an uncomfortable truth about a lot of nesting software: it doesn’t nest your parts. It nests rectangles. Every part gets an invisible bounding box drawn around it, and the software packs the boxes. An L-bracket, a ring, a gusset with a big notch cut out of it — as far as the nester is concerned, they’re all just rectangles of empty air, and all that air is sheet you’re paying for.
When I started building the nesting engine for PolygonLogix, I decided that whatever else it did, it would nest the actual shape. That decision leads you straight to something called the No-Fit Polygon, and it’s worth understanding, because it’s the difference between a nester that saves you material and one that just arranges rectangles quickly.
What a No-Fit Polygon Actually Is
Take two parts. Slide one all the way around the other, keeping them touching but never overlapping — like tracing one cookie cutter around another on the counter. The path the sliding part’s reference point traces out is the No-Fit Polygon, the NFP. It’s a complete map of every legal position: outside the NFP, the parts can’t touch; on it, they’re kissing; inside it, they’d collide.
Once you have that map, placement stops being trial-and-error collision checking and becomes geometry. The engine computes NFPs using Minkowski sums — the mathematically exact way to combine two shapes’ boundaries — so a concave part can genuinely tuck into another part’s notch, and a small part can nest inside the cutout of a bigger one. That’s where the density comes from: an L tucked into an L, rings stacked inside rings, the stuff bounding boxes can never do.
Kerf and Spacing, Baked Into the Math
Part spacing isn’t a rule the engine checks after placing — it’s inflated directly into the NFP geometry itself. If you set 0.25″ spacing, every no-fit boundary grows by exactly that much, once, before placement ever starts. Parts can then legally sit right at the limit, as tight as your kerf and your material’s heat tolerance allow, with no safety margin stacked on top of a safety margin. (Getting that “once, not twice” right cost me more debugging nights than I’d like to admit.)
Pattern Stamping: Where the Speed Comes From
Exact NFP math is dense but historically slow — early versions of this engine took two or three minutes on big jobs, and one memorable build took 157 seconds on a benchmark that now runs in about three. A pile of the speedup came from classic engineering: precomputing NFPs in the background across every CPU core while you’re still setting up the job. But the biggest single idea was pattern stamping.
When you’re cutting 500 of the same bracket, the engine doesn’t need to solve 500 unique placement problems. It solves a small one — what’s the densest repeating arrangement of this part against itself, using the exact NFP — and then stamps that pattern across the sheet like a die. Full true-shape density, at a speed that feels like cheating. The engine also runs multiple placement strategies against each other and keeps whichever nested tighter, and it’s smart about rotations: a circle doesn’t need 360 orientations tried, a rectangle needs two, a highly irregular part automatically earns 45° steps, and a small part headed into another part’s cutout gets tried every 15° to squeeze in.
The Nest Isn’t Done When the Parts Are Placed
Most nesters stop at first placement. TurboNest treats that as a rough draft. After everything is placed, compaction passes take over: a guided local search slides parts against each other to squeeze out slack, a “jostle” pass runs the slide in reverse to shake parts into gaps the first direction missed, and a sequence-refinement pass swaps the order parts were placed in, re-nests, and keeps the result only if it’s tighter. None of these ideas are mine — they come from two decades of published nesting research (Burke, Egeblad, Gomes & Oliveira, Dowsland), and the engine implements them the way the papers describe. The difference is you don’t have to read the papers; you just get the sheet back.
Arcs Stay Arcs
One more thing that matters at the machine: the engine is arc-aware end to end. A hole in your DXF is treated as a true arc through nesting and stays a true arc in the output — so your machine gets real G02/G03 moves, not a 200-segment polygon pretending to be a circle.
Why You Should Care: Utilization Is Money
Every percentage point of utilization is sheet you don’t buy. On a job running eight sheets of half-inch plate, the gap between rectangle-packing and true-shape NFP nesting isn’t academic — it’s often the difference of a whole sheet, on every repeat of that job, forever. And because PolygonLogix quotes from the actual nest, the material cost in your quote is the material you’ll really use — which means you can price sharper than the shop across town who’s quoting off a spreadsheet guess, and still keep your margin.
That’s TurboNest. The screenshots on the features page show it packing 500 parts across 8 sheets — that’s a real nest, not a mockup.