On Jul 7 2013 6:19 AM, Matt Shaver wrote: > On Sat, 6 Jul 2013 20:23:12 +0000 > Chris Morley <chrisinnana...@hotmail.com> wrote: > >> I guess it really comes down to at what performance machine does >> jerk >> limitation show real benefits ? > > I would guess that almost all machines _we_ deal with would benefit > from jerk limiting.
when I first read this I thought "here we just wrote of 99.9999% of the world," but then I read that you found it an interesting discussion. So, I assume you did not mean to stifle the discussion but to give a reality check. That's OK, but I would like to explore the outer boundaries and ask the question -- if LinuxCNC was up to the task, how far can we push it? I for one have on my list of things to do (maybe even this fall) rebuild a CNC plasma torch with a 16'x25' bed (yes those are feet, not inches) and the gantry alone weighs an estimated 1/2 to 3/4 ton. Oh, I forgot to mention that the original controller moved the beast over 500IPM, and the drive motors are several KW each... I also know that there are actuators out now that can move 10m/s with acceleration profiles over 20g, and if you are moving short ranges I've seen a voice-coil that could accelerate at 200g's! I do not want to limit myself to thinking of how to retrofit my #16 Van Norman, but what could I do if I built a delta-tau machine and needed it to move well over several meters a second? Would LinuxCNC be able to eat that elephant? > A machine that might be able to get away without > limiting jerk would be something where the moving parts had very > little > mass, like maybe a tiny plotter or a tiny gimbal. > >> And I still don't see why one would want to turn off jerk limiting >> for some machine movements. > > Here's what I think Jon is talking about: When in a rigid tapping > cycle, the Z axis is "slaved" to the rotational position of the > spindle. The spindle will have its own characteristics of motion, but > we don't always have tight servo control over them. For example, at > the > bottom of a tapped hole, the spindle must reverse, but it may come to > a > stop with a jerk due to high friction (you've probably experienced a > tapping operation where the tap wants to move in small jerks rather > than smoothly cutting). At this point the Z axis should probably be > controlled with no limit on jerk (and maybe no acceleration limit > either) so as to maintain the spindle-to-Z-axis position relationship > which is critical since they are now _mechanically_ interlocked > through > the tap and it's threads in the workpiece. From this I see a software engineering option/directive -- either allow the acceleration profiles and/or jerk minimization to be switched on/off at individual g-codes. I can also see this as a high level option to turn it on and off for both testing and as needed. I would then suggest that we buy a handful of taps and push the machine to its limits and break a few. Up to this point we have not discussed at all how to implement jerk minimization. For years I've wanted to experiment with using a NURB as the low level representation (instead of Catmull-Rom spline I thought it used). We might be able to use CAGD smoothing techniques on the NURB (or other spline representation) and solve it a priori geometrically. >> Interesting discussion. > > Yes! ------------------------------------------------------------------------------ This SF.net email is sponsored by Windows: Build for Windows Store. http://p.sf.net/sfu/windows-dev2dev _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
