Another Jerk (and more!) controlled controller

[Mooselake]

I stumbled across recent TinyG item today discussing their approach to jerk (along with snap, crackle, pop, and an as yet unnamed 7th order property) controlled motion.

TinyG is software for a juiced up (double clock rate) arduino type usb cnc controller.  They also have a port to an ARM based Arduino Due called TinyG2.

I don't know enough to tell if the extra orders are significant or just marketing, but it sounds interesting.

Kirk
. 

[ArtF]

Thanks Kirk:

  I enjoyed that. Written by someone that knows the math by the sound of it. As to marketing or not?
Marketing!. lol, even Auggie with 6th order is too high according to some conversations Ive had.
Greg, from Tormac once told me he figured anything above 4th order would be unnecessary on
most machines, and he may have been right, but over the years I had many responses to Tempest, so
I know it helped out quite a few.  3rd order is jerky, snap order removes 90%, crackle another
8% , pop another 1.5%..by the time you get to the unnamed 7th order (Crispie?), I doubt
one could see it in any realistic sense. Seems a waste or processing time as it raises the
complexity of the calculations. But then..if you have the CPU time.. , never say never. I suspect
its like 1 million count encoders, more marketing than useful.

  I say congrats to them, nice to see the field progressing.

Art

[Mooselake]

So snap, crackle, and pop aren't just something that Synthetos developed, then?  I was wondering if they were still around, they haven't released any hardware or done any Github updates for a couple years.

Crispie, crunchy, creepy for the next 3?

I don't remember if it was here or another forum, but I clicked and watched a link to an hour long Youtube video by the developer of LinuxCNC's latest planner.  It sort of indirectly suggests that a gcode preprocessor could do at least some optimization and reduce the burden on real time code, but I haven't really found any.  Is there some reason they haven't caught on, at least for 2.5D code?

Kirk

[ArtF]

That video is pretty much how Auggie works, but Auggie's planner does go further to snap, crackle and pop correction.

  Im in the middle of the module that processes Gcode into shapes for reprocessing, so we'll see
over time how much sense that makes..

Art

[chicoxiba]

TinyG's planner is Exact Stop AFAIK. There is no constantVelocity or segment blending.. so speed would be handicapped highly if I'm not mistaken. But here is a very impressive video of tinyG running a bit of 3d printing gear..https://www.youtube.com/watch?v=Om0wTqFA-Dw

Regards,
F

[ArtF]

Thx for the update.. I was wondering how they crammed all that in a small cpu.. makes much better sense now.

(Thats thing screams.. )
Art

[chicoxiba]

just found this and thought it would be worth sharing..

cheers, F

[ArtF]

Hi:

Thx for the paper, in a nutshell, its basically how Auggie works..

Art

[chicoxiba]

Hi Art

Just found another cool paper...  http://research.engr.oregonstate.edu/mp ... alCopy.pdf

have fun

cheers,
c.

[ArtF]

Hi:

  Interesting paper. Its basically Auggies planner again in almost every way, the single exception being the blend formulas,
in that I use a hermite curve rather than a bezier. The two share the most important qualities though. Im quite happy
with Auggies motion generally, though I have tweaked a few things recently in acceleration computations. I will be doing
more work on it this winter as I burn more with my laser and I'll see what improvements I can make. Auggie has few users for
motion, of course lasers via windows isnt a large field of people, and cnc is now a very different animal than when I wrote
Mach3, so I dont suspect Auggie's motion controller will ever grow to a large enough community to expand it too much. I
do have a galvo scanner Id like to add to it soon though so you never know what capabilities will show up in there. I add
anything I write to the release of course, so anything could show up. If you find motion bad in any particular way always let me
know so I keep it back of mind for when I open that code next.
  I know your troubles are usually surfboard related if I recall right, so speed at small segment is  the type of trouble you
hit.  Thats a very typical result on that type of work, I think a more bezier motion may help , some sort of nurbs capability to get rid
of all the small line segments.. or elliptical motion capability rather than arc.  Anyway, keep sending such papers,I do like to
keep up ono current thinking..

Thx
Art