On 10/24/2013 02:32 AM, Robert Ellenberg wrote: > Hi All, > I've made some progress on the trajectory lookahead features. To test > the idea of blending with arc segments, I modified the code to be able to > follow a tangent path without slowing down. There's no blending done in the > code yet, but it's able to follow G-code program that's been pre-made to > have only tangent lines and arcs. You can see the difference in speed in > the videos here: > > http://www.youtube.com/playlist?list=PLTi4ghj9Dyag612Q75Gg8eTfhwrq1PvHu > > For 0.004" segments on a machine with a max speed of 1.8 in/s and a max > acceleration of 15 in/s^2, the difference is pretty dramatic. > > The code is available in my github mirror of linuxCNC. This code has > *ONLY*been tested in simulation, and may crash or behave unpredictably > if you > don't use the pre-optimized G code given below. > > Here is the branch with my current changes: > https://github.com/robEllenberg/linuxcnc-mirror/tree/feature/arc-blend-demo > > To see arc-blending in action, build this branch with simulation enabled, > and load/run the following G-Code file: > > nc_files/random_blended_0.004in_1.8ips_15ips2_22-Oct-2013.ngc > > To see a direct comparison of this run with the original blend method, > checkout this branch instead: > https://github.com/robEllenberg/linuxcnc-mirror/tree/feature/arc-blend-demo-original > > Load and run this file to see the original parabolic blends on the same > path: > > nc_files/random_*unblended*_0.004in_1.8ips_15ips2_22-Oct-2013.ngc > > The next step is to add a routine in the tpAddLine function that creates a > blend arc segment between a previous line, and inserts it into the queue. > Once that's working, I'll add the optimizer to allow it to speed up and > slow down safely. > > At this point, I'm curious if anyone can see any roadblocks or unintended > consequences of these changes. For reference, here's the more in-depth > writeup on this method that I posted a few days ago: > > https://www.dropbox.com/s/o5he7ijpplkvc8r/Trajectory%20Lookahead%20with%20Arcs.pdf > > Thanks! > Rob > ------------------------------------------------------------------------------ > October Webinars: Code for Performance > Free Intel webinars can help you accelerate application performance. > Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from > the latest Intel processors and coprocessors. See abstracts and register > > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk > _______________________________________________ > Emc-developers mailing list > Emc-developers@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-developers > Wow, Thanks for this work! I am used to writing tangential paths, as I spent many years with AGie Wedm's. Their control insisted on tangential paths. Tangency simplified the UV tracing the XY path, and allowed offset paths with no more/less segments. At one time, I wrote a blend radius changer ( 'Blech' :). It allowed radically different tool diameters for rough and finish passes. Another tool reversed the path allowing chaining for multiple passes refining the path to finished dimension. You just cylced thru D values cutting closer and closer to the final path compensation. ( see IchGuchLive's recent YouTube post ... thx Ich! ) regards TomP
------------------------------------------------------------------------------ October Webinars: Code for Performance Free Intel webinars can help you accelerate application performance. Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from the latest Intel processors and coprocessors. See abstracts and register > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
