And here’s an update: Turns out I’m wrong after all regarding Trapezoid power!
I asked in the Lightburn forum if the offset fill job setup could be modified so that at sharp corners I get overscan. This led to a bit discussion, which finally let me understand how things work - I seem to have been exeptionally obtuse here not understanding this earlier:
- Constant Power Mode means laser is started with M03 - and this actively switches OFF trapezoid/s-curve power
- Non-Constant power mode starts the laser with M04 - and this activates the power compensation
Inline power is just independent from that.← It is a beast of its own, and I need to test this separately. It is neither M03 nor M04, but the G1 command include Sxxx - and if this properly uses trapezoid power I need to test - later…EDIT: See below - no trapezoid power with inline power
EDITEDIT: It does! Combining M04 with Inline power does do power adaption! 
Now comes the mean part which made me despair: Lightburn Snapmaker profile has a bug - regardless if I select constant power mode on or off in Lightburn job setup, it always issues M03, i.e. constant power in the GCode! And that lead me to believe trapezoid power is faulty in the Snapmaker implementation. I apologize for my wrong accusation!
Some people say (incl. Lightburn support): If you have issues with the Snapmaker profile, use GRBL profile, works just fine (it, e.g., addresses the problem with Air assist). So I tried GRBL - but this then has a problem with my much loved offset fill job setup - the laser turns off after the first move.
Long story short: I will do three things:
- Try to convince Lightburn to correct the Snapmaker profile for this plus air assist
- As long as the bug is there run a search’n’replace over Lightburn GCode to replace M03 with M04
- Reveisit this test here, as my interpretation of the results may suffer from the Lightburn bug: Laser Inline Power processing slow?
EDIT: I guess I should also revisit PCB etching with the 2W… would suffer from the same bug… Huh, lots to do the next weeks…
What remains is the problem with the 10W. This also is visible with Luban generated GCode.
Finally, here’s a picture of my tests:
- Top row - unmodified Snapmaker profile output: First: Square, offset fill, constant power OFF, next: Square, offset fill, constant power ON, next: Square line fill, constant power OFF, 5mm overscan
- Middle row: Same setup, but GRBL processor
- Bottom row: Same setup, Snapmaker processor, for second and third square manually replaced M03 with M04 → Success!
