Look down into a part before the top layers close it and there is a clean channel running around the inside of every perimeter, with the infill stopping just short of it. On a finished part you cannot see it, which is the problem: it is a structural defect that is invisible by the time anyone inspects the piece.
Why a small gap matters more than it looks
Perimeters carry most of the load in a printed part, and they do it best when the infill is genuinely tied to them. A part whose infill merely sits near the walls behaves like a thin-walled shell with loose packing inside: it feels solid, weighs the right amount, and buckles at a fraction of the force.
It also makes the top surface harder to close, because the skin now has an unsupported run at every edge where the wall meets the infill.
Why the two paths do not meet on their own
Infill and perimeters are separate extrusions, planned separately. The slicer places the infill line's centre a set distance from the perimeter's centre, and expects the two beads to overlap by their widths.
Three things break that expectation. If the overlap parameter is set low, the beads are placed too far apart to touch. If flow is under-calibrated, both beads are narrower than the plan assumed and the intended overlap evaporates. And if the perimeter has already cooled below welding temperature when the infill arrives, the two touch without bonding, which leaves a joint that is geometrically closed and mechanically open.
That third case is the one people miss, and it is why print order and speed matter here at all.
What to set
- Infill-to-wall overlap around 25% of line width. This is the direct control. Higher closes the gap harder but starts pushing the wall outward, which produces infill telegraphing through the surface — the two defects sit at opposite ends of the same dial, and the correct value is where neither appears.
- Enable an infill anchor length if the slicer has one. It extends each infill line a short distance along the perimeter before turning away, which welds the joint over a length rather than at a point. Where this option exists it is more effective than raising overlap.
- Print walls first, then infill. The infill is then laid against a wall that is still warm and can be pressed into.
- Narrow the speed difference. Infill running several times faster than perimeters arrives with less heat per unit length and has less time to weld. Bringing the two closer together costs a little print time and improves the bond directly.
Calibrate before you compensate
If flow is under-set, raising overlap masks a general shortfall in delivered material — and the same shortfall is quietly thinning every wall on the part. Check the wall thickness on a single-wall test cube first; if it is under nominal, fix that and re-slice before touching overlap at all. See under-extrusion for how to tell a flow shortfall from a melt-rate limit.
Pattern choice makes a difference
Some infill patterns meet the wall at a favourable angle and some do not. Lines and grid patterns approach the perimeter obliquely and produce short, weak contacts. Patterns that run a boundary path of their own, or that meet the wall closer to perpendicular, tie in better. Concentric infill has no gap problem at all by construction, at the cost of being poor in every other respect.
Gyroid at low density is the common offender: its contact with the wall is intermittent and curved, so a value that works with a grid can leave visible channels here.
Voids that are not this fault
If the void is inside the solid region rather than at the wall, that is under-extrusion or a partial clog rather than a placement issue. If the gap only appears on the sloped parts of the model, the infill is meeting a wall whose position moves layer by layer, and the fix is more solid layers near sloped surfaces rather than more overlap.
Materials worth mentioning
PETG shows this more than PLA because it shrinks slightly more as it cools and pulls away from the perimeter after being laid against it. Polyamides do the same thing to a greater degree, which is one reason nylon parts benefit disproportionately from a warm chamber. PLA in a well-calibrated profile rarely shows it, so a sudden appearance on PLA after a slicer update is worth checking against the overlap default, which several slicers have changed over the years.