I have been experimenting with PPI using LinuxCNC (formerly EMC2). The very non-intuitive relationship between speed and PPI had me perplexed at many times. After reading more info on the Bart's experiments the idea of "power delivered" and "linear energy density" and their relationship to feed rate, PPI, and laser power was more clear.
"Power delivered" is a measure of how many watts is being pushed into the material, watts = joules / sec, which is a per second quantity. PPI times feed rate determined this value. More PPI or more feed rate increased the delivered power. Which is opposite of how most view the relationship, faster feed rate should cut deeper.
"linear energy density" is a measure of how long the "power delivered" spends at each linear point. This determines how deep a cut is. If one inch is cut at a constant power delivery level, how long this one inch takes determines how deep the cut because 1 second per inch will be half the total power send into the same linear distance as 2 seconds per inch. To go deeper PPI is increased and feed rate is decreased by the same proportion.
After all this I decided to simplify to feed rate and duty cycle
. To determine the "power delivered" all you need to do it multiply duty cycle by laser power level (ie 40w at full power). To get "linear energy density", just set feed rate. The end result if if the material is too melty, reduce duty cycle. To cut deeper increase feed rate.
I have implemented this in LinuxCNC using Ben Jackson laserfreq.comp. Using reverse spindle to mean duty cycle (ie M4 S55 is 55% duty cycle). The code https://github.com/jv4779/2x_laser