Back
[08:38:13] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15isladelobos opened issue #64: tkemc button help-version fails 02
https://github.com/LinuxCNC/linuxcnc/issues/64
[09:12:31] <cradek> uh-oh, we have a tkemc user
[09:16:42] <skunkworks_> http://forum.linuxcnc.org/forum/38-general-linuxcnc-questions/31043-show-axes-letters-instead-numbers#75389
[09:16:50] <skunkworks_> (same guy)
[09:53:57] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.6 f1f7b27 06linuxcnc 10tcl/tklinuxcnc.tcl tklinuxcnc UI: rebranding * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=f1f7b27
[09:53:57] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.6 407a46e 06linuxcnc 10scripts/rip-environment.in rip-environment: rebranding * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=407a46e
[10:01:43] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15SebKuzminsky commented on issue #64: Thanks for the bug report.... 02
https://github.com/LinuxCNC/linuxcnc/issues/64#issuecomment-223595830
[11:09:20] <linuxcnc-build> build #364 of 4017.5.deb-wheezy-armhf is complete: Failure [4failed shell_3] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/4017.5.deb-wheezy-armhf/builds/364 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[12:01:42] <seb_kuzm1nsky> zlog:
[14:18:31] <PCW> puzzled by "clearpath servo" thread
[14:18:32] <PCW> I just cannot duplicate the following error no matter what I do
[14:18:34] <PCW> (but I have not tried rotary axis just their really high accel on linear axis)
[14:18:35] <PCW> Their setup is a bit weird (rotary axis accel is extreme = 8000 degrees/s^2 so 16 ms to full speed but linear axis are slow ~150 ms to full speed)
[14:24:44] <skunkworks_> the fact he is getting a following error - doesn't it point toward a tuning problem?
[14:25:26] <skunkworks_> stepgen<>motion?
[14:26:23] <PCW> Except I have tried the same numbers and see no problem (and even wildly different numbers) and cannot duplicate his large following errors
[14:28:16] <PCW> Tormach did have the thread order messed up but I think just about everything looks right now and its still broken
[14:28:26] <skunkworks_> With the same stepgen acc/vel overhead?
[14:28:38] <PCW> yes
[14:29:26] <PCW> I tried the same setup (8000 accel and 9600 stepgen accel and 125 velocities and see no issue at all)
[14:30:03] <PCW> (well other than the accel slope is in about 18 velocity steps)
[14:31:06] <skunkworks_> sim runs g1z3a3600f120 just fine with the rotory settings like his acc/vel
[14:31:48] <PCW> OK so I'm stumped
[14:31:50] <PCW> PathPilot bug?
[14:32:12] <skunkworks_> no clue - have not played with it yet.
[14:33:04] <PCW> Just making sure its not a stepgen/PID issue
[14:33:43] <PCW> pretty hard to debug without halscope
[14:35:36] <PCW> Their are definite issues with their hal file but this sort of looks like something else
[14:36:38] <skunkworks_> I am pretty sure pathpilot still has halscope. I think it is something like from mdi - admin halscope
[14:36:57] <PCW> I even had the guy look at the card jumpering sine the only way I could get that wild behavior is having a crazy wrong stepgen clock
[14:41:38] <cradek> do you have the whole ini and hal config?
[14:43:07] <skunkworks_> http://forum.linuxcnc.org/forum/pathpilot/31037-pcnc770-4th-axis-w-clear-path-servo-pathpilot?limitstart=0#75368
[14:43:13] <skunkworks_> it is tormachafied
[14:44:07] <skunkworks_> I could setup a 7i90 config with those setting just to see. (it would be a good exersize.)
[14:44:07] <cradek> what happened to his FF1?
[14:44:30] <cradek> and I thought you told him to stop using MAX_ERROR
[14:45:39] <skunkworks_> further on he changes max_error - but I think you may have spotted it?
[14:45:47] <skunkworks_> FF1?
[14:45:58] <cradek> it was 1 on cnczone, now it's 0 which is surely wrong
[14:46:27] <cradek> oh I see page 2 now
[14:46:31] <cradek> grr forum
[14:46:37] <skunkworks_> in his paste - it is also missing ff1
[14:47:01] <cradek> yeah I bet he changed it mistakenly
[14:47:23] <skunkworks_> Do you want to post?
[14:50:12] <skunkworks_> Probably a silly question.. ;)
[14:51:55] <cradek> hey I even know my password
[14:51:56] <cradek> I'll post
[14:52:10] <skunkworks_> Great! Thanks for noticing that.
[15:28:06] <PCW> Thanks Cradek, maybe he's been changing one thing at a time and reverting the previous change...
[15:28:57] <cradek> it's so easy to get confused when trying a bunch of things (and so frustrating when helping someone when that happens)
[15:29:33] <cradek> it's good, now that he gets immediate ferrors when it's wrong
[15:33:49] <PCW> yeah much better than making things that are machined wrong
[15:37:04] <PCW> I have to remember to not have people set the stepgen FERROR limits too small
[15:37:05] <PCW> as that does not work without the DPLL (you need to add servothread jitter*maxvel without the DPLL)
[15:42:34] <PCW> I can get below about 200u degrees peak error with the DPLL and decent tuning (thats about 1/25 of a step in his setup)
[19:21:18] <seb_kuzminsky> jepler's famous (by accident this time):
https://lwn.net/Articles/689792/
[19:24:05] <andypugh> What are the chances of that?
[19:24:55] <andypugh> (actually not such a trivial computation, it’t the Birthday thing with an unknown sample size).
[19:25:50] <andypugh> ie, a class of 30 children is very likely to have two with the same birthday, because they already share a year, and 1 in 365 * the combinations is small
[19:38:05] <cradek> jeez those keys we all use are old
[19:40:05] <cradek> guess I signed jeff's but he didn't sign mine
[19:43:08] <cradek> pcw: I don't think that guy's twisted part is programmed with A motion other than indexing, is it? I don't see how it could end up twisted.
[19:44:01] <cradek> so weird
[19:46:42] <PCW> I dont understand either but showing how the gcode runs might help diagnosis
[19:47:47] <PCW> At least I know its not some weird stepgen thing
[19:49:22] <skunkworks__> zlog
[19:50:44] <PCW> I learned a lesson: to make sure I say change A,B,C all at once, I got the impression he changed one thing at a time and undid the previous change each time
[19:51:55] <PCW> (or maybe just edit the files and post the edited files)
[19:58:17] <skunkworks__> I bet if he tries to line up at 360 mark after the program run (just a marker) it doesn't line up.
[19:58:47] <andypugh> skunkworks__: Yes, that does seem like the next thing to be sure of.
[19:59:04] <andypugh> 1024 counts poer rev not 1000?
[19:59:19] <PCW> yeah i kind of suspect step polarity (or scaling maybe 1000/1024)
[19:59:25] <PCW> ha
[19:59:37] <andypugh> Twist and slow settling might be two unrelated problems.
[19:59:46] <cradek> yeah bet so
[20:00:02] <PCW> the slow settling is fixed
[20:00:33] <andypugh> Yes, I saw that. I am suspecting that the slow settling was never the problem
[20:00:35] <cradek> I bet it's spindle noise making it turn while it's supposed to be still
[20:01:13] <andypugh> It was the obvious wierd, but would not show up in the parts.
[20:01:49] <andypugh> Took 5 seconds to settle? how much metal is removed in 5 seconds?
[20:01:58] <PCW> yeah a step/dir servo probably has pretty high bandwidth step inputs (and not really suited to singe ended drive)
[20:02:04] <andypugh> (And, really, is there money in making artisinal dice?)
[20:02:29] <skunkworks__> lots of D&Ders out there
[20:02:43] <skunkworks__> never got into that. cousins did
[20:03:01] <skunkworks__> now there are online games that take the stats to the nth degree
[20:03:09] <andypugh> Never tried it, always thought I ought to.
[20:05:00] <seb_kuzminsky> andypugh: my kids are super in to it, we play sometimes
[20:05:00] <andypugh> Wierd fact: I have been credited (almost certainly inaccurately) with writing the first computerised wargame. But I have never played a conventional tabletop wargame.
[20:39:09] <jepler> eek (pgp thing)