A support engineer can usually tell within a minute whether a complaint is about the machine or about the file. The symptoms differ in a specific way: a machine fault is inconsistent, while a file fault is perfectly consistent — it does exactly the wrong thing, every time, in exactly the same place.
That distinction matters commercially. A shop that blames the machine for a file problem stops production, waits for a technician, and often pays for a visit that ends with someone opening the file and clicking one button. A shop that recognises the pattern fixes it in ten minutes.
This article covers the ten things that actually go wrong between a design and a cut. It is written in the order a file travels — from the drawing, through export, into the machine's software, onto the bed — because that is the order in which errors accumulate.
The most common complaint in laser cutting, by a wide margin, is that the part came out the wrong size.
Almost always the cause is units, and almost always it is one of three things.
The file was drawn in one unit system and imported as another. A drawing in millimetres read as inches comes out roughly twenty-five times too large; inches read as millimetres comes out correspondingly too small. The factor is so large that it is instantly obvious — which makes it easy to fix but also easy to be embarrassed by.
The design was made at a reduced scale. Architectural and furniture drawings are often produced at a ratio for printing. If a part is drawn at one fifth of full size, the machine will cut it at one fifth of full size, faithfully and without complaint.
The export applied a scale factor. Some exporters convert units on the way out; some preserve the number and change the label. A file that opens correctly in one program and wrongly in another is almost always this.
The defence is simple and worth institutionalising: put a known dimension in every file. A reference rectangle of a stated size, or a dimension line on the artwork, gives everyone — designer, operator, and any technician — a way to verify scale in one measurement. When a file arrives without one, measuring a known feature before cutting the whole sheet is the cheapest test in the workshop.
A vector that does not close is not a shape; it is a line that stops. The machine will trace it faithfully and leave the part attached to the sheet, or produce a gap that nobody notices until assembly.
Open paths come from three sources. Breakpoints left by editing — a path split to delete a segment and never rejoined. Traced artwork — auto-tracing often produces outlines that look closed at screen resolution but have microscopic gaps. Imported drawings where two segments share a position but not a node.
The check is to look at the nodes rather than the shape. Zoom to the outline, select it, and look for endpoints — a closed path has no loose ends. Most software can also report whether a selected path is closed.
Two practical notes. Fill the shape and see whether it holds — an open path often fills unpredictably, which is a fast visual test. And close paths deliberately rather than dragging the last node onto the first: a dragged node sits close, not coincident, and the gap returns as a hairline in the cut.
The most wasteful file fault is also the most invisible: the same line existing twice.
It happens through copy-paste over an existing outline, through offsetting a curve and deleting the original badly, through merging two versions of a drawing, or through auto-trace producing a doubled contour. On screen it is indistinguishable from a single line.
The consequences are all bad and all avoidable: the beam cuts the same path twice, so the kerf effectively doubles, the edge chars more, the job takes longer, and on thin material the part can fall apart or distort.
Detection is a matter of discipline rather than eyesight. Select all and count objects — a shape that should have one outline and reports two has a duplicate. Look at the cut preview for a doubled path — many controllers draw repeated moves as a heavier line. Weld or union the artwork before export, which removes exact overlaps and joins touching outlines.
A related fault is near-duplicate geometry: two lines a hair apart, from an imprecise trace or a careless offset. Those do not merely double-cut; they can produce a sliver of material that breaks loose during cutting and damages the work or the bed.
Some software interprets an outline's stroke width as a real width and cuts two parallel paths — one on each side of the stroke — instead of a single centre line.
The result is a slot rather than a cut: the part is undersized by the stroke width on every edge, and on fine detail a small feature can disappear entirely.
The fix is to use a hairline or zero-width stroke for anything that will be cut, and to keep decorative stroke widths for artwork only. The trap is that a drawing can look correct on screen — a heavy line looks like a confident outline — while being wrong for a machine that reads geometry rather than appearance.
A file containing live text depends on the font being present on the machine's computer. When it is not, the text either substitutes — wrong spacing, wrong shapes — or fails to render at all, and some systems cut a placeholder box.
The rule is unambiguous: convert text to curves before export. Not after the file is sent, not in the machine software, but in the design file, before it leaves the designer.
Two consequences worth knowing. Converting to curves makes the file impossible to edit as text later — so keep the editable original alongside the cut file, named clearly. And converted text increases node count substantially, which brings the next problem.
A single file usually carries more than one intention: cut the outline, engrave the logo, score the fold line, mark a reference. The machine distinguishes them by layer or colour, mapped to operations in its software.
That mapping is where good files become bad parts. A logo intended as an engrave cuts straight through the panel because the colour was mapped to cut. A fold line intended as a score cuts the part in two. The reverse — a cut line mapped to engrave — leaves a panel still attached to its sheet, which is discovered when the operator tries to lift it.
The discipline is a small standard, applied every time: a fixed palette for a fixed meaning. One colour for cut, one for engrave, one for score, one for marks, documented and used by everyone who produces files. Files from outside designers arrive with their own colours, so the layer mapping has to be checked rather than assumed.
Auto-traced artwork and over-detailed vectors can carry thousands of nodes along a contour that needs a fraction of them. The cut is correct and the machine crawls: every node is a change of direction, and the controller spends its time decoding rather than moving.
The symptoms are recognisable. The head hesitates along curves. A job that should run smoothly takes far longer than its length suggests. Curves show tiny facets under magnification rather than a clean arc.
The fix is to simplify the path — most vector software offers a smoothing or reduction function — but with care. Over-simplification changes the shape, and on parts that fit together that change is a tolerance error. Simplify the decorative contours; leave the functional geometry alone.
A "vector" file can contain a bitmap. Logo artwork pasted from a web page, a scan, a photograph placed in the drawing — each arrives as pixels, and the machine either ignores it or, if traced, produces the ragged outline of every antialiased edge.
The check is to look at the object list rather than the canvas. A bitmap appears as an image object; a true vector appears as paths. If the job needs the logo cut, the artwork must actually be vector, not merely look like it.
Two of the faults above combine into a complaint that reaches every support desk: the job took far longer than expected, and the edge is over-charred.
The usual pair is duplicate geometry and excessive node count — one doubling the path, the other slowing the motion. Neither is a machine problem, and neither is visible on screen at normal zoom. Both are found by selecting the geometry and reading what the software reports about it.
Shops ask which format to send. The more useful answer is that format matters less than discipline, because the common faults cross format boundaries.
Still, there are practical differences worth knowing:
The practical recommendation for a shop receiving files from customers: agree one format, ask for a known dimension inside the file, and check the layer mapping on arrival. Those three habits prevent most of what follows.
Before a sheet is loaded, a short routine catches nearly every fault described above.
The last row is the one that makes the rest optional in practice: cut one part, measure it, and check the fit before committing the sheet. Two minutes against a wasted panel is the best trade in the workshop.
Nine of those ten are file problems. That ratio is the reason this article exists: the machine is usually doing exactly what it was told.
A laser is an obedient machine. It cuts what the file describes, at the size the file states, along the paths the file contains, in the operations the layers dictate. Every one of the faults in this article is the machine being correct and the file being wrong.
That is good news, because file faults are cheap. They are found by measuring one known dimension, looking at nodes rather than shapes, counting objects, and cutting one part before cutting a sheet. None of it requires a technician, and all of it can be standardised into a two-minute routine.
The shops with the fewest "machine problems" are not the ones with the best machines. They are the ones whose files arrive in a state the machine can obey.
Hot News2026-11-16
2026-11-12
2026-11-10
2026-11-09
2026-11-06
2026-11-05