How to Draw with a Laser Cutting Machine: A Star Micronics Fabricator's Honest Mistakes

After seven years of handling CNC machining and laser cutting orders at Star Micronics, I have personally made 23 significant mistakes that added up to roughly $18,000 in wasted material, rework, and missed deadlines. The most expensive part of custom fabrication isn't machine time or tooling. The most expensive part is the information gap between what you designed and what the operator actually needs.

When I first started, I assumed the machine was the bottleneck. I thought a better fiber laser or a newer CNC would fix our quality problems. It didn't. The real bottleneck was in the drawing file, the material notes, and the assumptions I didn't write down.

How to draw with a laser cutting machine? The short version

If you are asking 'how to draw with a laser cutting machine?', here is the answer you actually need: work in a vector format at 1:1 scale, keep all cutting paths on one layer, and use a separate color for engraved or etched lines. That is 80% of it.

• Vector file, not raster. Use .dxf, .svg, .ai, or .eps. A PDF can work if text is converted to outlines, but a high-res PNG is not a cutting file.
• 1:1 scale. The file must be actual size. I have personally sent a part to production that was exactly 10 times too small because the drawing was in the wrong unit. Seriously, check the units.
• Closed paths. Every path needs a clear start and end point. An open path causes a weird lead-in or a skipped segment.
• Layer discipline. Put all cut lines on one layer and all score/engrave lines on another. The machine uses line color or layer name to decide power and speed.
• Kerf is real. A fiber laser removes material. Typical kerf in our laser tube cutting setups runs roughly 0.006 to 0.012 inch depending on material thickness. For tight slots, adjust geometry by half the kerf per edge.

Bottom line: a pretty PDF is not a cutting file.

Vector files are not optional: the star micronics logo lesson

We get logo requests every month. Someone will search for 'star micronics logo', download a PNG, and expect us to cut it. When I open that file, I have to ask for vector artwork. What most people don't realize is this: I can trace a PNG, but every trace is an interpretation, and the reproduced logo is almost guaranteed to be slightly wrong.

I went back and forth with a customer once about redrawing their logo because they didn't have the original artwork. I thought I was helping. The result: the dot on their 'i' was 0.4 mm off center, and they noticed before we did. That taught me the hard way that redrawing someone's logo without approval is a risk, not a favor. Now the rule is simple: if you can't provide vector art for a logo cut, we can't guarantee logo accuracy.

Laser tube cutting in Minnetonka: the file questions people forget

Laser tube cutting in Minnetonka is one of our most requested services, and it has the same root issue as flat sheet cutting: unclear files. The machine is fast, but it needs to know which end is which in the drawing. No, really—I had a 10-foot tube order cut with holes reversed because the drawing looked fine from the top view.

For tube cutting, add a datum note. Call out 'reference face' or 'front end'. If the tube will be bent after cutting, show a bend reference. It feels like overkill until you scrap a $700 bundle of stainless tube.

End mill flat end for stainless steel: choose geometry before coating

If you landed on a page titled 'end mill flat end for stainless steel', I will save you twenty minutes: the tool is rarely the problem; the setup is. A flat end mill for stainless steel needs a strong geometry, but you can make the best carbide end mill look terrible by running it at the wrong speed or with bad chip evacuation.

I'd argue that consistent chip thickness matters more than coating. Stainless steel work-hardens fast. If a flute rubs instead of cuts, the heat goes into the part and the edge, and the tool fails quickly. For a flat end mill in stainless, use a lower surface footage than you would use in mild steel, keep stepover modest, and use high-quality coolant or mist. A TiAlN-coated carbide end mill is a good starting point, but it won't fix a shallow cut that rubs.

The pre-flight checklist I use before every order

After the third rejection in Q1 2024, I created a pre-flight checklist. In the past 18 months, it has caught 47 potential errors. I only believed in this checklist after ignoring the unit check once and turning a $3,200 order into scrap. So yes, I check it every time now.

1. Is the file vector? Are text fonts converted to outlines?
2. Are dimensions in the units I expect?
3. Are all paths closed?
4. Are cut and engrave lines separated into different layers?
5. Is material type and thickness specified?
6. For stainless steel: will the tool see sheared edges or machined edges?
7. For tube: where is the front or datum?
8. What tolerance applies? If none, we use ISO 2768-m, which gives roughly ±0.2 mm for sizes from 6 to 30 mm. Reference: ISO 2768-1.

There are other checks, but these are the ones I missed when it mattered.

What's changed in five years, and what hasn't

What was best practice in 2020 may not apply in 2025. Five years ago, a 2 kW fiber laser was a big deal; now it is table stakes. Laser tube cutting and 5-axis CNC are no longer exotic. The fundamentals haven't changed, though: you still need a clear drawing, the right tool, and someone who knows how the machine will interpret the file.

When these rules don't apply

To be fair, the advice above is for typical sheet and tube fabrication orders. It doesn't cover thick-plate laser cutting with heavy heat input, where the heat-affected zone changes hole sizes. It doesn't cover 3D-printed parts where support removal matters. And if you are prototyping, a quick and dirty file is probably fine—you will learn more from a cheap part than from a perfect drawing.

If you ask me, the easiest way to avoid these issues is to talk to the shop before sending a file. Not because we expect your file to be perfect. Because after $18,000 of my own mistakes, I would rather spend five minutes asking about the front of a tube than five weeks explaining why a part doesn't fit.