very cool! as far as optimization depth - I am sure it depends on the machine and the cutting profile. Is that something that could be configurable? I had done some testing between mach and linuxcnc a while back. Running spiral.ngc - anything over about 9 or 10 segment look ahead didn't effect much.
Just to clarify - you are actually processing the profile further ahead - more than just 2 segments? (or is this strictly arc-blend vs parabolic?) sam On 11/4/2013 12:13 AM, Robert Ellenberg wrote: > Hi All, > > I've completed a beta version of circular arc blending, which looks pretty > promising! Here's a performance comparison between the stock blends and > circular arc blends: > > http://www.youtube.com/playlist?list=PLTi4ghj9Dyajbv4F4P4TxWXpgPgvqrhA- > > For a random, kinky path with short segments, the difference is pretty > dramatic. The average speed is about 3ipm with stock blends, vs. 7 ipm with > circular arc blends. > > Here's the branch on github: > > https://github.com/robEllenberg/linuxcnc-mirror/tree/feature/3d-blend-arc-techdemo > > The algorithm itself is working, so I'd like to make sure the > implementation is robust too: > > - Edge cases and checks such as minimum / maximum blend radius > - Eliminate any conflicts with other machine modes (haven't found any > yet, but I wouldn't be surprised if there were). > - How much difference does optimization depth make? Depending on machine > config, much more than 10 segments might not matter > - Parameters stored in INI file and passed in via commands > - Profiling / RT timing performance, especially on slower machines > (tpAddLine has a lot more work to do now) > > If you have ideas for additional features or tweaks, or would like to do > some testing, definitely let me know. In particular, I'm looking for > feedback on the following areas in simulation and/or hardware: > > - Are synchronized motions (CSS, rigid tapping) adversely affected by my > changes? I don't have a spindle pickup on the sherline, so this would be > very helpful to test this on a machine so equipped. > - Position accuracy at high speed, perhaps a comparison of G2/G3 arc vs. > a circle made of line segments. > - Profiling / timing performance on a real machine (simulation profile > results show no more than a few percent difference, but I'm not sure how > representative that is of hardware conditions) > - Qualitative performance on high-acceleration machines. Can the machine > hit speeds so that trapezoidal profiles aren't smooth enough? > > So far I've tested 3D motion, but it should be able to handle motion in all > 9 axes (by falling back to parabolic blends). I'd recommend running your > programs in simulation mode first just in case. > > -Rob > ------------------------------------------------------------------------------ > Android is increasing in popularity, but the open development platform that > developers love is also attractive to malware creators. Download this white > paper to learn more about secure code signing practices that can help keep > Android apps secure. > http://pubads.g.doubleclick.net/gampad/clk?id=65839951&iu=/4140/ostg.clktrk > _______________________________________________ > Emc-developers mailing list > Emc-developers@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-developers > > ------------------------------------------------------------------------------ Android is increasing in popularity, but the open development platform that developers love is also attractive to malware creators. Download this white paper to learn more about secure code signing practices that can help keep Android apps secure. http://pubads.g.doubleclick.net/gampad/clk?id=65839951&iu=/4140/ostg.clktrk _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
