Laser Engraver Settings by Material: Power, Speed and Pass Counts from 5W Diode to 60W CO2
Settings only transfer between machines when both numbers refer to optical watts at the lens, not wall-plug draw. Convert your machine's advertised figure first — a marketplace "40W" diode is usually a 5-10W optical module, and a K40-style "40W" CO2 tube typically measures nearer 30W new. Then take the power, speed, pass and interval values below as the centre of a test grid.
Take a setting written for a "20W" machine, load it into your "20W" machine, and there's a fair chance you scorch straight through a sheet of birch ply. The number on the box is the problem. Some makers quote optical watts leaving the module; others quote what the whole unit pulls from the wall, and on the worst listings the two differ by a factor of six.
Everything on this page is keyed to optical watts. Find your machine in the conversion table, note the real figure, then read across the row for your material. Speeds are in mm/min throughout — if your controller is a Ruida and reports mm/s, divide by 60.
None of these values are finished recipes. They're starting points, chosen to land you inside the useful window on the first grid rather than the fourth.
Convert the number on your box before you copy anything
Two conventions are in circulation and nobody labels which one they're using. Reputable diode brands now quote optical output and list input power separately, so a 10W module means roughly 10 optical watts from two combined 5W emitters pulling about 60W at the wall. Marketplace listings frequently do the opposite: they print the input figure, or an invented "equivalent", and the actual emitter puts out a fraction of it.
CO2 is messier again. A glass tube's rating is a peak figure, and Chinese tubes are graded generously — plenty of tubes sold as 80W measure closer to 70W when new and drop from there. Then the beam has to survive three mirrors and a focusing lens, each of which takes a small bite. Clean molybdenum or silicon mirrors sit around 97-99% reflectivity apiece and a coated ZnSe lens transmits about 97-99%, so a clean train delivers roughly 88-94% of what leaves the tube. Grubby optics can halve it.
What the box says What that number is usually measuring Typical wall-plug draw Optical watts leaving the source Watts at the workpiece, clean optics Plan settings around 5W diode module, named brand Optical output, one emitter 24-32 W 4.5-5.5 W 4.2-5.2 W 5 W 10W diode, two emitters combined Optical output 55-75 W 9.5-11 W 9-10.5 W 10 W 20W diode, four emitters combined Optical output 110-150 W 18-22 W 17-21 W 20 W 40W diode, eight emitters combined Optical output 220-300 W 34-42 W 32-40 W 38 W "40W" / "80W" / "130W" diode on a marketplace listing Input power, or an invented equivalent 40-130 W 5-12 W 5-11 W Ask for the emitter count, then assume 5 W each K40-style 40W CO2 desktop Tube's peak rating, usually optimistic 500-700 W 28-35 W when new 25-31 W 30 W 40W CO2, branded tube Tube rating at full duty 600-900 W 38-42 W when new 34-38 W 36 W 60W CO2 glass tube Tube rating at full duty 900-1300 W 55-62 W when new 48-56 W 52 W 80W CO2 glass tube Tube rating; many measure short 1200-1700 W 65-78 W when new 58-70 W 64 W 100W CO2 glass tube Tube rating at full duty 1500-2200 W 90-105 W when new 80-94 W 86 W Carry the last column into every settings chart you read, this one included. Wall-draw figures assume the chiller is on the same supply. If your machine falls between two rows, interpolate on optical watts, never on the advertised number. Tube age matters more than most people expect. A CO2 tube that's had a couple of thousand hours through it will deliver noticeably less than its rating, and it degrades gradually enough that you blame the material instead. That's exactly why a baseline test file is worth keeping — more on that below.
The settings matrix
Read down to your material, across to your machine. Each cell is power percentage, then speed in mm/min, then pass count. The right-hand column is scan interval, which controls how dark and how detailed a raster engrave comes out.
Diode entries assume a 445nm blue module with air assist running. CO2 entries assume a 2-inch focal lens and a working air assist nozzle.
Material 5W optical diode 10W optical diode 20W optical diode 40W CO2 (~35 W at work) 60W CO2 (~52 W at work) Interval / DPI Birch plywood 3 mm, raster 25% · 3000 · 1p 15% · 4000 · 1p 10% · 6000 · 1p 15% · 12000 · 1p 12% · 15000 · 1p 0.08 mm / 318 DPI MDF 3 mm, raster 30% · 3000 · 1p 20% · 4000 · 1p 12% · 6000 · 1p 18% · 12000 · 1p 14% · 15000 · 1p 0.08 mm / 318 DPI Basswood 3 mm, raster 20% · 3500 · 1p 12% · 4500 · 1p 8% · 6000 · 1p 12% · 12000 · 1p 10% · 15000 · 1p 0.08-0.10 mm Cast acrylic 3 mm, frosted engrave Clear: no mark. Black: 20% · 2000 · 1p Clear: no mark. Black: 12% · 3000 · 1p Clear: no mark. Black: 8% · 4000 · 1p 25% · 6000 · 1p 20% · 7200 · 1p 0.10 mm / 254 DPI Veg-tan leather 2 mm 20% · 4000 · 1p 12% · 5000 · 1p 8% · 7000 · 1p 12% · 15000 · 1p 10% · 18000 · 1p 0.08-0.10 mm Anodised aluminium, dye ablation 100% · 1000 · 2p 100% · 2000 · 1-2p 100% · 3000 · 1p Marking compound required Marking compound required 0.05 mm / 508 DPI Powder-coated steel tumbler 100% · 1200 · 2p 100% · 2000 · 1p 90% · 3000 · 1p 40% · 6000 · 1p 30% · 7200 · 1p 0.05-0.08 mm Slate coaster 100% · 1500 · 1p 80% · 2500 · 1p 60% · 3500 · 1p 60% · 6000 · 1p 45% · 7200 · 1p 0.10 mm / 254 DPI Clear glass, frosting No direct mark — coat first No direct mark — coat first No direct mark — coat first 25% · 6000 · 1p, damp paper 20% · 7200 · 1p, damp paper 0.10 mm / 254 DPI 300 gsm cardstock, light engrave 15% · 4000 · 1p 10% · 5000 · 1p 6% · 8000 · 1p 8% · 15000 · 1p 6% · 18000 · 1p 0.10 mm / 254 DPI Format is power % · mm/min · passes. Treat every cell as the centre of a 3x3 grid: bracket it 20% either side on power and 30% either side on speed, cut the grid on your own stock, and keep the tile. Three cells say a diode can't do it, and those aren't hedges. Blue 445nm light passes straight through clear cast acrylic and clear glass without depositing energy, so there's nothing for the beam to work on. You can cheat glass with a sacrificial coating — a coat of flat paint, or wet newspaper laid over the surface — but you're marking the coating and letting the heat conduct, which is a different process with different failure modes.
Anodised aluminium is the reverse case. Blue diodes ablate the coloured dye out of the anodic layer beautifully and leave a crisp silver mark, while 10.6 µm CO2 barely touches it without a marking compound. If tumbler and nameplate work is the plan, that's a genuine argument for the diode.
Interval, DPI and the setting most people over-cook
- mm/min is the GRBL default on diode machines; mm/s is the Ruida default on CO2. 6000 mm/min is 100 mm/s. Getting this wrong by a factor of 60 is the single most common cause of a ruined first job.
- DPI and scan interval are one setting written two ways: 25.4 ÷ DPI = interval in mm. So 254 DPI is 0.10 mm, 318 DPI is 0.08 mm, 508 DPI is 0.05 mm.
- Setting an interval finer than your focused spot adds no detail. It overlaps the passes and darkens the burn. A 0.10 mm spot has no business running a 0.03 mm interval — you're just tripling the job time and cooking the surface.
- Doubling DPI roughly doubles run time and roughly doubles the heat delivered per square millimetre. If a photo comes out muddy, drop the DPI before you drop the power.
- For photo work on wood, a dithered image at 254-318 DPI usually beats a greyscale image at 500 DPI on any diode machine. Greyscale relies on the module's low-power linearity, which most consumer diodes don't have.
Materials that never go in the machine
This list isn't about results. It's about what the smoke does to your lungs, your rails and your control board, and none of it is recoverable by changing a setting.
Material What it releases under the beam What it does to the machine Verdict PVC, vinyl, PVC-backed faux leather, foamed PVC board Hydrogen chloride gas, which becomes hydrochloric acid on any damp surface Corrodes linear rails, lens coatings, control boards and the extraction fan from the inside out Never. No setting makes it safe. ABS Hydrogen cyanide and a heavy, sticky smoke Melts instead of vaporising; films the lens within minutes Never. Polycarbonate / Lexan Dense smoke and open flame Absorbs 10.6 µm poorly below the surface, so it yellows, catches and drips Never for cutting, pointless for engraving. PTFE / Teflon Fluorine compounds, hydrogen fluoride at temperature Attacks optics, and at concentration attacks the operator Never. Chrome-tanned leather Chromium compounds, potentially hexavalent Nothing structural — the fume is the whole problem Never. Veg-tan only. Fibreglass, carbon fibre, epoxy laminate Resin fume plus abrasive airborne particles Particles score the lens and pack the bearings Never on a hobby machine. Polystyrene and polypropylene foam Styrene fume Flashes into open flame almost instantly Never. That's a fire, not a cut. Anything you can't positively identify Unknown Unknown Get the safety data sheet, or put it down. Check this list against every new material before the first test tile, not after. If a supplier can't tell you what a sheet is made of, that answer is itself the answer. Everything on the permitted list still needs fume extraction ducted outside. A filter box with carbon that's been saturated for six months isn't extraction, it's a noisy box.
Burn the grid before you burn the job
Five minutes here saves an afternoon. Most software ships a material-test generator — LightBurn calls it the Material Test array — and it does the whole thing for you.
- Set the grid to sweep power across the columns and speed down the rows, centred on the chart value for your machine and material.
- Use a 10 mm square per cell with a 2 mm gap. Bigger cells waste stock; smaller ones don't show heat build-up.
- Run the grid on an offcut from the same sheet you'll use for the job. Ply from two different pallets behaves differently, and that isn't superstition — glue line depth and species vary batch to batch.
- Label the tile with the machine, the lens and the date, in the laser, on the tile. A drawer of unlabelled test tiles is worthless within a month.
- Re-run the grid whenever you change lens, change air assist pressure, or fit a new tube. Not because the chart changed — because your delivered watts did.
Where this chart will be wrong for you
The matrix assumes a clean lens, a working air assist and a workpiece sitting flat at the correct focus height. Take any one of those away and the numbers slide.
- Plywood glue lines. A 3 mm birch ply with two glue lines needs noticeably more energy than solid basswood of the same thickness, and the lines sit in different places sheet to sheet.
- Moisture. Stock that's been in an unheated garage over winter will steam rather than cut, and every value here will read as underpowered.
- Air assist pressure. A few litres per minute of clean air across the nozzle changes cut depth more than a 20% power increase does. Too much air on wood feeds the flame instead.
- Lens condition. A ZnSe lens with a light haze can cost 10-20% of delivered power without looking obviously dirty.
- Bed height. Being 1 mm off focus with a 2-inch lens costs roughly two-thirds of your power density. It costs almost nothing with a 4-inch lens.
- Tube hours. A well-used 60W tube may be delivering 40W. Your settings didn't change; your machine did.
Three things that move results more than power ever does
Focus first. Every hour spent on a repeatable focus routine pays back more than any amount of power tuning, because power density falls off with the square of the spot diameter and a 0.5 mm focus error is easy to make and invisible to the eye.
Air assist second. It clears the plume out of the beam path and stops smoke absorbing the light you already paid for. On acrylic it also decides whether the edge comes out flame-polished or frosted and pitted.
Material consistency third. Buy your ply and acrylic from one supplier in reasonable quantity and settle your settings against that stock. Chasing a moving material with a fixed chart is the reason so many people conclude their machine is faulty when it's working exactly as designed.