Troubleshooting a laser, by symptom

Most problems are one of five things: focus, dirty optics, the material, the file, or something loose. Work in that order — it goes from most likely and free to least likely and expensive. Changing power first is the common mistake, because it hides the cause instead of finding it.

Updated 2026-09-16

Find the symptom, check the causes in the order given. They are ordered by how often each turns out to be the answer, not by how obvious it seems.

It will not cut through

1. Focus. Set from the top of the material, and re-set whenever thickness changes. This is the answer more often than everything else combined.
2. Dirty lens. A haze you can barely see absorbs a third of your power. Clean it before you touch a setting.
3. The material. Voids or dark glue lines in cheap plywood stop the cut in patches. If it fails in one spot and cuts elsewhere, suspect the sheet.
4. Wavelength. A diode will never cut clear acrylic at any power — it is transparent to blue.
5. Not flat. A bowed sheet lifts part of the cut out of focus. Weight it down.
6. Only now: add a pass, then raise power.

The finished part is the wrong size

1. Kerf, if it is out by a fraction of a millimetre. Cut on the line, a part measures a kerf less than drawn. Compensate, or size the drawing for it — not both.
2. Units, if it is out by a lot. An SVG with no units is read as pixels; at 96 dpi a 100 mm square arrives around 26 mm.
3. Steps per mm, if everything is out by the same percentage. Cut a 100 mm square, measure, and correct the axis calibration.
4. Belt tension, if one axis is wrong and the other is not.

The engraving is offset, or lands on the wrong part

1. Operation order. Engrave and score must run before cutting. A part cut free can shift, and anything engraved afterwards misses.
2. Origin. Confirm where the job thinks zero is, and that it matches where the machine thinks zero is.
3. Lost steps. If the offset grows through the job, something is slipping — belts, or acceleration set too high.
4. Backlash. Run the same shape twice in one job; if the second does not land on the first, there is play in the drive.

Edges are charred, or there is a brown halo

1. No air assist. The single biggest factor on wood. Smoke in the kerf absorbs the beam and the flame licks the surface.
2. Too slow. Prefer more passes at higher speed to one slow hot pass.
3. Mask the surface with transfer tape; it takes the staining instead of the wood.
4. Material. Resinous woods scorch whatever you do.

Corners are darker than the straight sections

The head decelerates into corners, so at fixed power more energy lands per millimetre exactly there. Use GRBL’s M4 dynamic power mode, which scales power with speed, and set a sensible minimum power so the beam does nothing while the head is crawling.

Circles are not round, or shapes do not close

1. Belt tension, and both belts at the same tension.
2. Loose pulley grub screw on a motor shaft — a classic, and it produces a flat spot in one axis only.
3. V-wheel preload. Wheels should turn under light finger pressure but not rock.
4. Acceleration too high for the machine, which rounds corners and overshoots.

It will not connect

1. Browser. Web Serial works in Chrome and Edge, not Firefox or Safari.
2. Another program holding the port. Only one application can own a serial port at a time.
3. Cable. Charge-only USB cables carry no data, and look identical to ones that do.
4. Driver, on some boards — CH340 and CP2102 chips often need one.
5. WiFi machines: check signal at the machine, inside its metal enclosure, not at the desk.

It connects, then drops out mid-job

1. USB power management putting the port to sleep. Disable selective suspend for that port.
2. Ground loop between machine and computer. A USB isolator is cheap and often fixes it outright.
3. Interference from the laser power supply or a stepper driver. Route the USB cable away from them.
4. WiFi drop-outs. Give the board a fixed address so it does not change mid-session.

The machine refuses to move

Usually Alarm. A board with homing enabled starts there deliberately, because it does not know where the head is. Home it, or clear the alarm if the current position is acceptable. If it alarms again immediately, a limit switch is triggered or wired inverted.

It worked last month and does not now

1. Dirty lens, overwhelmingly. Smoke builds gradually, so the decline is gradual and looks like a dying source.
2. Different sheet. "3 mm ply" varies between batches in thickness, species and glue.
3. Focus, if anything about the setup changed.
4. Genuinely ageing source — last, not first. Diodes fade slowly; CO2 tubes usually fail more abruptly.

The order that saves time

1. Focus — free, thirty seconds, most likely.
2. Optics — cheap, five minutes, second most likely.
3. Material — try a known-good offcut.
4. File — check units and operation order.
5. Mechanics — belts, wheels, grub screws.
6. Settings — a test grid, not a guess.
7. The source — last, because it is the expensive answer and rarely the right one.

Written by a human with AI assistance and reviewed before publication. Mistakes are still possible — treat settings as starting points and test on scrap.