Leslie Newell wrote: > I have finally got to the point on my lathe where I can start tuning. > The setup is: > Mesa 5i20 driving a 7i29 with 1024CPR encoders on the screws. The PWM > gain is set so the velocity is roughly equal to the commanded input. For > example a PID output of 60 will give about 60mm/sec movement. > > I have the P and D set up to be just under the oscillation point and > following error is fairly low. Adding I increases stiffness when the > machine is stationary but also increases the following error when > moving. Even more strangely, the following error keeps increasing on > long moves. It is almost as if I is working the wrong way round while > moving. If I turn down I, the following error while moving is much less > and it does not increase on long moves. Adding FF1 also increases the > following error. So far the best results so far are with P=1200, D=10, > everything else 0. Has anyone got any idea what is going on here? > Because of discontinuities is the programmed path, the I term is of limited usefulness. I usually just put a tiny amount in and leave it there. I can't give a fixed value, as it needs to be proportional to P. With P=1200, then I would expect I to be about 10. If increasing FF1 makes following error worse, reduce it. FF is not a feedback, so it will easily "overdo" the commanded velocity or acceleration. FF2 requires extremely small values, I often use something around 0.005, but you can suppress error spikes on acceleration very well with it.
Assuming 5 TPI screws directly driven, then 1024 CYCLE/rev encoders produce 20480 counts/inch, which is a bit low for a system without velocity feedback. See my VERY well hidden article on servo tuning at http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?PWM_Servo_Amplifiers Jon ------------------------------------------------------------------------------ _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
