On Tue, Apr 10, 2007 at 09:49:26PM +0800, Xuecheng Xi wrote: > Dear Developers, mostly to Cradek: > > The trajectory command for a circle is not really a circle. The higher > feedrate, the bigger the radial contour errors are. > > When I run the G-code (feedrate=100mm/s, radius=40mm) > G2 x40 y0 i-40 j0 f6000, > > I record the trajectory comands rx and ry from > axis.0.motor-pos-cmd > axis.1.motor-pos-cmd > > then I calculate the contour error, i.e, > error=sqrt(rx*rx+ry*ry)-40 > > I found the reference input is a circle with a slightly smaller > radius than the desired value (radius=40mm). Although these errors are > quite small for low feedrate (feedrate=20mm/s, error=0.16 micron), > these errors are quite big for high feedrate (feedrate=100mm/s, > error=4 micron). We can imagine, with bigger radius and higher > feedrates, these errors will be even bigger. As the circular reference > inputs are calcuated from cosine and sine function, they should be > "perfect" with respect to the "float" type storage resolution in C. > > Not sure whether anyone else found this. It would be great that > someone will help to fix this problem. > > Best regards > Xuecheng
Hi, can you give more information about how you sampled these positions? Are they sampled every traj period, or every servo period, and what is the ratio between traj and servo? If traj != servo, you're getting cubic interpolation on the output of the planner itself. If you set traj = servo, I think you will get exactly the planner output. Also you say it's not really a circle, but on your error plot I see a straight line, which if I understand the plot, means the path is circular but too small. Can you clarify that please? Chris ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
