ons 2022-12-21 klockan 13:23 +0000 skrev andy pugh: > On Wed, 21 Dec 2022 at 12:12, Stuart Stevenson <stus...@gmail.com> > wrote: > > > > A point and a vector fully describe a plane in 3D space. > > Indeed, and you could certainly have a command to set a generic > plane. > But you don't need the point. G17 and friends do not define a plane, > they define a continuum of planes (for example G17 is all XY planes > for every possible Z) > > There are two discussions here that are getting mixed up. (and > neither > correspond to the subject any more) > > 1) Arbitrary arcs. There are a few ways to do this. You could > certainly define an arbitrary plane then run the arc. With a normal > vector defined the G2/G3 direction is well defined. But you could > also > specify start, end and centre (in 3D space) on one G-code line which > also fully describes two possible arcs. Or (start) end and waypoint, > using (XYZUVW) and (IJK) which unambiguously defines a single arc.
Specify three points in 2D or 3D space: start, centre and end points as you suggest I think is a good idea that will work for an arc. However not totally sure specify centre point should be totally disregarded. If I think correct path should be the same in the three dimensions XYZ if arc is in XYZ space only or if tool direction UVW also change. For a foam cutter I expect "tool tip" is decided to be on a certain point on the wire while UV change direction. If foam cutter is wire between two axis moving in XY direction I suggest kinetamics have to resolve it into XYUV to make programming same as 5-axis mill, however not sure which coordinate system is used for programming a 5-axis mill. Expect arc work also for example for only XY coordinates if nothing more is needed. Left out axis is assumed constant at current position. Arc will be circle for one turn instead of a spriral but if I think correct spiral could be programmed with several half circles. Not sure adding an extra spiral option is good or confusing. > Defining an arbitrary plane would mean that spiral arcs remain > possible, but there is quite a lot of maths in defining it. End > points > and waypoint can be pulled off of a CAD model. > There is an old "arbitrary arcs" branch, > https://github.com/LinuxCNC/linuxcnc/tree/arbitrary-arc that uses > magic comments to define the arc. But I think this was just a proof > of > concept and the plan wouldn't have been to keep that method. > > 2) Simultaneous curves (arcs, NURBS) for foam cutters. This would > need > a way to either define two end points and two centres on one G-code > line, or a way to "cache" end points and centre on the line before > (which would be unusual but not entirely outside the G-code > paradigms). > Maybe PQR for UVW centre? PQ would certainly be possible for UV arcs, > re-using R would be a bit unusual for UW and VW arcs. > This wouldn't work for NURBS, though, which already uses P. But > perhaps Q could be used for the "secondary" plane in that case. > Consistency is a problem. If I think correct it should be possible to program a foam cutter the same way as a 5-axis mill. A point on the wire is assumed to be the "tool tip" while kinematics solve the rest of the problem. Nicklas Karlsson _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
