#emc-devel | Logs for 2007-06-22

[00:16:17] <skunkworks> I think the wiki method does a pretty good job - but I have no experience. Just looked agian and I had calculated it wrong - used Kc for the i and d calc instead of the Kp. which seems to make it even better. be cool to try it with load attached.
[00:27:27] <skunkworks> with the emc doc method I get following errors of over .003. With the wikipedia method I get following errors less than .0003
[00:28:19] <jepler> someone should fix the emc docs then
[00:29:45] <skunkworks> well - I would feel more comfortable if someone like cradek tried that method on his machine.. To see how it performs under real world applications.
[00:30:16] <skunkworks> or maybe rayh
[00:31:39] <skunkworks> with the emc docs I can turn the shaft probably 1/4 turn while the wikipedia method I cannot turn it more than a few tenths of a thousands
[00:32:00] <skunkworks> until I knee over the servo supplsy
[00:32:03] <skunkworks> supply
[00:32:57] <skunkworks> quite neat
[00:44:55] <cradek> with your favorite tuning, do you get much overshoot or ringing if you step position?
[00:45:41] <skunkworks> step position?
[00:45:52] <cradek> I don't trust ZN or similar on my lathe, because the oscillation is so small that it's within backlash and no mass is really moved
[00:46:07] <skunkworks> oh - I forgot about that.
[00:46:17] <skunkworks> (backlash)
[00:46:20] <cradek> or a small incremental jog even
[00:46:46] <cradek> yeah it oscillates maybe 1/16 turn of the screw (.06 mm)
[00:47:17] <cradek> at least that's my guess why it doesn't seem to work
[00:51:33] <skunkworks> You have not spring loaded your nuts yet? :)
[00:52:10] <skunkworks> cradek: do you have halcope pictures of your tuning?
[00:57:38] <cradek> http://timeguy.com/cradek-files/emc/manual-final.png
[00:57:40] <cradek> maybe this
[00:58:10] <cradek> the error (.006) is mm
[00:58:21] <cradek> this is probably an incremental jog (so full speed)
[01:00:50] <cradek> actually it doesn't look so great - it stops about .006mm away, then takes 100msec to settle
[01:01:30] <cradek> it's probably because of the dead spot around zero of the pwm drive
[01:01:53] <cradek> notice how the I windup doesn't do anything to help for almost 100msec
[01:03:59] <skunkworks> ah huh. :)
[01:04:33] <cradek> could play with that for days, and not make parts any better :-)
[01:04:48] <jmkasunich_> hi guys
[01:05:11] <skunkworks> Hi
[01:05:21] <cradek> hi
[01:05:25] <skunkworks> is the pid error the same as the axis following error then?
[01:05:30] <cradek> yes
[01:05:32] <skunkworks> ok
[01:13:20] <jmkasunich_> jmkasunich_ is now known as jmkasunich
[01:21:16] <jmkasunich_> jmkasunich_ is now known as jmk2
[04:35:45] <jmkasunich_> jmkasunich_ is now known as jmkasunich
[05:02:22] <jmkasunich> not good... the 5i20 seems to lose counts when turning backwards
[05:11:19] <jmkasunich> 10000 counts between index pulses going forwards, about 9700-9800 going backwards