The red beam is off centre — what that actually tells you

The red guide beam is not the cutting beam. It is coaxial with it, which makes it the fastest way to check the whole optical path without cutting anything.

If it is visibly off centre, diverging, or irregular in shape, stop here. Check the optical path before anything else — before focus, before gas, before you open the parameter table.

Reading the mark on the tape

A clean test: punch a mark on tape at the nozzle and look at where it lands.

• Mark concentric with the nozzle bore — nozzle and beam path are aligned. Move on.
• Mark off centre — the head is out of alignment. This alone will produce a bevel, a one-sided cut face, or an unstable process.

This matters most after a rebuild. The single most common mix-up is a full-reflection mirror fitted where the deflecting mirror belongs. Everything looks assembled correctly and nothing cuts properly.

When alignment is fine and the alarm is still there

If the mark is concentric and you still get faults, the trigger is the process, not the mechanics. A useful split:

• Close the shutter and run the job with the beam off. If the alarm appears anyway, it is motion-induced — typically the capacitance alarm triggered by the machine's own movement rather than by cutting.
• If it only appears with the beam on, the trigger is in the cutting process itself, and the alignment work is genuinely done.

That one test separates two problems that otherwise look identical on the screen.

The lens and the focus

Two things worth ruling out before you blame the lens:

• A contaminated focusing lens combined with a beam that has drifted off axis produces a slower, worse version of a pure alignment fault. Clean or replace, then re-check the red beam.
• On some heads the focus drifts quickly. The usual cause is an unreasonable motorised-focus parameter rather than the optic at all — replacing a perfectly good lens will not change it.

The order that saves time

Red beam first, then lens, then focus, then parameters. It costs a few minutes and removes the two most common false trails — swapping hardware that was fine, and tuning a parameter table that was never the problem.