Okay, I got this. With a budget of 20K per, all I need is some programming help. That Gcode seriously can not be easy to make. I would guess you make the part like a multi color part, then just choose which process for what piece. With 5 axis though a lot of that has to be collision detection.
ROS has planners for that. They work well for any A to any B (including collision avoidance and orientation), but I think you’d have to still come up with a list of points you wanted to move the spindle. I can conceptually understand how to do that with a 3D CNC, but I don’t know how to do that with 5DOF.
Making sure the tool is always normal to the surface (perpendicular on all axes). I could not imagine the kind of coding that would take, managing acceleration.
Marlin has not had an update in 5 days and there are all sorts of problems with the planners, delta, and corexy configs. Jerk did nothing for a few years and had tens of thousands of eyes on it, no one caught it. Delta accels don’t seem to be factoring in each stepper just the head, the planner is disconnecting the last segment of each move. Drop in 2 more rotational axes and that stuff would be a nightmare. Hats off to these guys for what looks to be like a solution, unless they just used some off the self commercial software. That could explain 10k of the 20k price.
The software doesn’t get much more complicated with more complicated math. It does get harder to understand. If you just split the problem up into pieces, each one can be tested, debugged and understood individually.
That said, there is a sort of mantra in software. It takes about 2x more understanding to debug software than to write it. If you make the software as clever as you, you won’t be able to debug it. So you’ll have fewer people that can do the 5DOF planner and debugging, but there are smart people out there writing papers on it. Implementing it isn’t much different than programming an Android app or Marlin. IDK, I’m sure it will happen someday, if it hasn’t already on some corner of the Internet.
So what I was seeing in the video was a standard 3 axis printer for the cylinder, then they rotated the bed 90° and switched Y out for a rotation, still printing in 3 axis. I imagine they can then rotate the bed to switch to normal Y and rotating X for the fifth axis. But that would still be 3 axis printing. Going to be a nasty, nasty slicer though to decide how to divide the job up into x,y,z, x,w,z, v,y,z sections…
Interesting. So you think it is sort of broken up, only any 3 axes (or 4) working at once? That would make things a bit easier. A hybrid. Interesting concept that could be possible.
That could be possible, just like we swap out an axis for a rotation physically, why not with software…hmmm.
Say we were to add an axis that just…flipped/turned something over…and started another job. We can just add a flip gcode in between jobs right, manually? With the dual axis firmware we have a really accurate home. Home, cut, turn (90/180/2…whatever), home, cut, repeat! Multi sided milling for the masses…Super easy for smaller jobs.
Except for that printed cap on the dome. That had to spin the dome on its base axis to clear the rest of the dome, kinda like spiraling up from the base to the top.
Found it, and yea, that’s going to be a bitch. Hmm, it still did a “normal” line pattern for the dome.
The dome looks like they were replacing x and y with two rotation axes. But they would also have to get the extrusion computed right. The amount of extrusion would need to know it was printing on a sphere or a flat surface.
These two demos do seem like hacks though, and they really should have printed a giant octopus if they really had the slicer figured out, so I’m guessing they don’t have a solution. It looked like they did have a solution for the milling, they were cutting neat parts with that.
There was a this old Tony where he added a 4th axis, and it was just to mill from another side (pi/2 rotations 
).
One big challenge is that it would only for for small things, because the z has to be as big as x and y.