> > Am 21.10.2009 um 01:11 schrieb Andy Pugh: > >> 2009/10/20 Haberler Michael <[email protected]>: >> >>> I dont fully understand the trajectory planner, but my gut feeling >>> it's a control loop oscillation which gets excited with the encoder >>> quantization/noise spectrum gets too close to the loop frequency. >> >> The problem seems to be with a bad initial velocity estimate. Chris
looking at axis.2.f-error it would occur to me it's not just the *initial* estimate here's a halscope shot of severe Z axis oscillation: http://mah.priv.at/gallery2/main.php?g2_view=core.DownloadItem&g2_itemId=70522 Note the following error is linearly increasing until some line of code kicks the joint into higher gear which makes it reduce f-error at a similar rate. When f-error gets close to zero, the process reverses. Interestingly the maxima of f-error become slightly less over time, while the resiudal errors increase. From the plot it looks like the tp has only a limited range of discrete speed settings - for the first five f-error maxima, like 'too fast' and 'too slow'; after the sixth f-error maximum it looks like it tries with a slower speed to correct the error and figures that's too slow after all. I admit I could be reading coffee suds or an artefact in the first place. I hope I hit the relevant signals to plot after all - hints appreciated. >> Radek came up with a patch that gave me a huge improvement with my >> 50ppr bit-of-laserprint-wrapped-round-the-spindle "encoder" >> >> I suggest you give it a try, though it does involve compiling your >> own >> development version of EMC2 at the moment. >> >> patch: >> http://timeguy.com/cradek-files/emc/0001-Improve-initial-threading-synchronization.patch in fact I did my tests with that particular patch applied, as well as with the patch in git; it seemed to help a bit. But clearly it's not a stable situation. I'll retry tomorrow with vanilla, git and this patch applied and see if it makes a difference wich can be quantified with my aural analysis procedure ;-) -Michael ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
