Jon Elson wrote: > I have no idea how hard it is to do better with this lookahead,
It's hard. :-( > I expect it is not easy to handle all possible combinations of > moves, with compensations being turned on/off, etc. But, for > people doing contouring of surfaces, they often want to skim > rapidly across surfaces at 60 IPM with several thousand > vectors/inch, therefore, at least 1000 vectors/second is needed. > This doesn't require large acceleration, as the vectors comprise > a smooth curve. If there is a discontinuity, then the machine > has to slow down, of course. That's the rub - if there is a discontinuity the machine has to slow down. But it doesn't know there is a discontinuity until it gets there. (Or in EMC's case, until it gets within one segment of there.) Multi-segment lookahead is _very_ non-trivial. We've discussed it at length several times over the years on IRC. There is only so much you can to in the realtime code - it needs to have roughly constant execution time, we can't be doing things that are order O(N) where N is all the lines in the program. You can in theory do all that lookahead in user space. But things like the user changing feedrate override on the fly, or probing moves, or several other issues, can completely mess up a nicely pre-planned path. There have been some intermediate steps - for example jepler wrote some code that detects collinear lines and merges them into a single line. It also detects and merges lines that are not collinear, but are close enough that if you merge them you remain within the tolerance that was specified by G64 Pwhatever. I'm not sure if that feature is in the 2.2 branch or only in CVS (to be released as part of 2.3.0). When you start considering arcs as well as lines it gets much hairier. Regards, John Kasunich ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
