#emc-devel | Logs for 2009-11-16

[00:06:55] <jepler> somebody should make comp error in that case
[00:23:35] <jt-plasma> jepler: thanks it is past that now :)
[00:25:06] <jt-plasma> back to our regularly scheduled errors now
[00:30:47] <jt-plasma> I would if I knew how
[00:33:58] <jt-plasma> all righty now we are back where we started :)
[01:23:14] <cradek> http://timeguy.com/cradek-files/emc/0001-fix-index-homing-thump.patch
[01:23:32] <cradek> jepler: jmkasunich: how about this? untested except in sim.
[01:46:02] <SWPadnos> cradek, isn't there also a step change in the D (feedback) term?
[01:47:47] <cradek> D term is derivative of error, not position; I think there is never an incorrect error because motion calculates it properly in spite of the position jump
[01:48:18] <cradek> there is ff2 (2nd deriv of command) and I think my patch also fixes that
[01:48:24] <SWPadnos> oh hmm - right, the error changes by + and - "the homing difference" at the same time
[01:48:40] <SWPadnos> yes, you'd get 0 for FF2
[01:48:47] <cradek> yes I'm pretty sure error is fine
[01:48:50] <cradek> right
[01:49:44] <SWPadnos> I was thinking that something could be done by looking at how the command and feedback values change, but that gets hard to think about pretty fast
[01:50:00] <SWPadnos> hmmm. there's a millipede thingy on my wall
[01:50:04] <SWPadnos> it might be time to clean my office
[01:50:51] <cradek> yeah you could try to guess when the position jump has happened, using a variety of methods - but I think knowing for sure is best, and we have that information
[01:51:32] <SWPadnos> yeah. I'm thinking about a more generic case where HAL is being used but EMC may not be
[01:52:00] <cradek> initial test - still gives a thump
[01:52:07] <cradek> seems less bad though, working on a plot
[01:52:26] <SWPadnos> the command has still changed, so the thump is probably in the next cyclenext
[01:52:30] <SWPadnos> -next
[01:53:32] <SWPadnos> hmmm. no, that shouldn't be it
[01:53:44] <cradek> hm, still one period of -10.00
[01:53:55] <cradek> I didn't hook it up
[01:53:57] <cradek> I'm a moron
[01:54:00] <SWPadnos> heh
[01:54:13] <SWPadnos> and you even changed the hal file in the commit :)
[01:54:20] <cradek> of servo-sim
[01:54:24] <SWPadnos> yep
[01:55:53] <cradek> PERFECT
[01:56:10] <SWPadnos> cool
[01:56:29] <cradek> feels perfectly smooth
[01:56:30] <cradek> yay
[02:01:34] <CIA-40> EMC: 03cradek 07master * rb7a2c6181d08 10/ (configs/sim/servo_sim.hal src/hal/components/pid.c): fix index homing thump
[02:06:28] <cradek> funny - I still cringe, waiting for it, every time so far
[02:07:13] <cradek> Z was really bad - the whole machine and floor shook/rang
[02:07:32] <cradek> funny that one ms can do that
[02:07:47] <SWPadnos> high current servos
[02:07:47] <cradek> I'm kind of surprised the amp doesn't fault when it happens
[02:07:57] <SWPadnos> shows the benefit of jerk limiting, I think
[02:08:19] <Jymmm> that's what she said
[05:09:21] <cradek> should emc-commit messages pretend they're From: the devel list? People frequently reply to them, and they come to me as <git-adm@git.linuxcnc.org>
[05:09:21] <cradek> or I could just add a Reply-To:
[05:49:40] <elson> elson is now known as jmelson
[05:49:53] <jmelson> Hello, all!
[05:50:16] <jmelson> Anybody here?
[10:20:36] <piasdom_> g'mornin all
[10:39:17] <piasdom_> skunkwork; are you up ?
[10:39:57] <piasdom_> *s;
[11:11:11] <piasdom_> where do i set the home location for g28 ?
[11:44:39] <jthornton_> I don't think it matters where you set it
[11:45:38] <jthornton_> oh he left
[13:11:31] <christel> [Global Notice] Hi all, we're having some problems with services right this moment -- we're looking into it. Apologies for the inconvenience.
[13:14:53] <services.> Services are presently running in debug mode, attached to a console. You should take extra caution when utilizing your services passwords.
[13:19:30] <alex_joni> jthornton_: he asked in #emc aswell
[13:45:24] <micges_work> odd, my mesa 5i20 hang up and don't want to boot up
[13:46:29] <micges_work> [26654.491610] hm2_5i20.0: FPGA did not reset: /INIT = 1, DONE = 1
[14:05:47] <micges_work> I have this many times, don't have clue why
[14:06:11] <SWPadnos> from reboot or from power-up of the PC?
[14:07:31] <micges_work> after power up pc it works fine once, and when I want to start emc again, it shows this error
[14:07:58] <micges_work> earlier if I had this error I swap mesa to another pci
[14:08:29] <micges_work> today I tourn off machine power for 10 min and it works
[14:08:41] <SWPadnos> well, I don't know what the problem is, but if there's a BIOS option to "reset ESCD data" or similar, you might try that
[14:08:46] <SWPadnos> or ECSD or something
[14:09:17] <micges_work> on earlier hang ups turning off power doesn't worked
[14:09:30] <micges_work> SWPadnos: thanks I'll try that
[14:09:35] <SWPadnos> and/or set the BIOS to "non-plug and play OS isntalled", if it has that option
[14:09:55] <SWPadnos> it may have nothing to do with that, since the card (usually) works on bootup though :)
[14:10:10] <micges_work> yep
[14:33:01] <micges_work> In hal file, net thc-cmd-original gui.thc.pos-cmd => hctrl.thc-cmd-in works
[14:33:08] <micges_work> but net thc-cmd-original gui.thc.pos-cmd =>hctrl.thc-cmd-in
[14:33:18] <micges_work> doesn't
[14:33:46] <micges_work> (doesn't work without space after '=>')
[14:39:04] <jepler> yeah, the "=>" has to be a separate word according to halcmd's simple tokenizer, which is based on whitespace
[14:53:51] <micges_work> ah ok
[15:04:18] <christel> [Global Notice] Hi all, Sorry for the services flapping, I think I have pinpointed the relevant module but not the actual crash (gdb with trace leaves me with 25K lines to look through!) -- to cause less disruptance I have unloaded the affected module, there will be no noticeable changes for end users, only staff usage of services. Apologies for the inconvenience and have a good day.
[15:31:24] <piasdom_> g'mornin all
[15:32:36] <jepler> hi piasdom_
[15:34:08] <piasdom_> did i do something wrong in #emc?
[15:34:18] <piasdom_> hello jepler
[15:38:11] <piasdom_> piasdom_ is now known as piasdom
[15:39:18] <jepler> beats me
[15:39:45] <jepler> hm, my real job is calling me .. bbl
[15:46:22] <piasdom> later
[16:43:43] <mozmck_work> does AXIS not require python-opengl?
[16:59:47] <alex_joni> mozmck_work: iirc it uses togl some small opengl implementation
[16:59:59] <alex_joni> which gets compiled into an python extension
[17:02:43] <mozmck_work> I see. I built a livecd based on 9.10. the first computer I tried it on AXIS gave some error related to opengl.
[17:03:05] <mozmck_work> That computer was way slow though and the next one it seemed to work fine.
[17:03:35] <alex_joni> if AXIS doesn't work, try glxgears
[17:03:53] <alex_joni> that's unrelated to emc2 and checks if GL works
[17:04:17] <alex_joni> usually you get an error about missing GLX extensions for your X server
[17:04:33] <mozmck_work> glxgears worked fine on that computer. my latency was only about 6us running from the livecd on that slow computer with glxgears running.
[17:07:13] <mozmck_work> but the computer I just tried it on EMC worked fine, so it may have had something to do with the other computer and how slow it was loading things.
[17:10:02] <alex_joni> ah, maybe
[18:18:16] <skunkworks> cradek: Nice job on the homing thump. (I think a few people have had that issue)
[18:18:37] <skunkworks> * skunkworks is finally on the mend.
[18:18:56] <alex_joni> The Movement for the Emancipation of the Niger Delta ("MEND") is one of the largest militant groups in the Niger Delta region of Nigeria.
[18:19:57] <skunkworks> hmm - did I spell that worng?
[18:20:00] <skunkworks> wrong
[18:20:08] <alex_joni> I doubt that ;)
[18:20:59] <skunkworks> heh
[23:54:41] <cradek> bizarre that nobody had complained about homing-thump, but it has been bothering people
[23:55:02] <cradek> I hope the way I fixed it is OK with everyone, because it's quickly getting entrenched
[23:55:15] <cradek> (I feel a little out of my league hacking on pid)
[23:58:42] <SWPadnos> PID seems easier than all that jommertry you're always working on :)
[23:59:24] <cradek> yeah but it's jmk's pony
[23:59:56] <cradek> the guts of emc, where I usually play, he couldn't give two shits about :-)