#linuxcnc-devel | Logs for 2014-10-22

[00:23:16] <cradek> argh, jogging is absolutely full of units bugs
[00:23:23] <cradek> how could we not have noticed this before
[00:24:42] <cradek> also the Vel: readout is wrong for rotaries when you switch the gui units
[00:26:28] <cradek> it's too late for this
[00:55:58] <seb_kuzminsky> linuxcnc-build: force build --branch=2.6 0000.checkin
[00:55:59] <linuxcnc-build> build forced [ETA 57m07s]
[00:55:59] <linuxcnc-build> I'll give a shout when the build finishes
[00:58:28] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.7 7261c9e 06linuxcnc 10scripts/githelper.sh Merge remote-tracking branch 'origin/2.6' into 2.7 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=7261c9e
[00:58:53] <seb_kuzminsky> there's still the problem with the v2.7.0-pre1 tag i made
[00:59:23] <seb_kuzminsky> i need to make a new release key -- the 2.6 key says 2.6 in the name, the new one will just say "LinuxCNC Release Manager"
[00:59:49] <seb_kuzminsky> then i'll put that key in the 2.7 keyring and make a 2.7.0-pre2 tag, and it should work
[00:59:52] <seb_kuzminsky> maybe
[00:59:52] <seb_kuzminsky> we'll see
[00:59:54] <seb_kuzminsky> good night!
[01:50:47] <linuxcnc-build> Hey! build 0000.checkin #2550 is complete: Success [3build successful]
[01:50:47] <linuxcnc-build> Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/2550
[09:42:51] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.7 ac7039e 06linuxcnc 10gnupg/pubring.gpg new release key for 2.7 and later * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=ac7039e
[09:42:51] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.7 c002147 06linuxcnc 10VERSION 10debian/changelog LinuxCNC 2.7.0~pre2 * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c002147
[09:42:52] <KGB-linuxcnc> 03Sebastian Kuzminsky 05signed tags 9033b2d 06linuxcnc 03v2.7.0-pre2 LinuxCNC v2.7.0~pre2 (tagged commit: c002147) * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=9033b2d
[10:01:31] <seb_kuzminsky> i think i did all the things to the buildbot so it wont freak out about 2.7 any more, we'll see in 45 mins or so
[10:02:41] <seb_kuzminsky> i updated the notes in the ReleaseCheckList wiki page so maybe i'll remember next time (http://wiki.linuxcnc.org/cgi-bin/wiki.pl?ReleaseCheckList)
[10:25:47] <jepler> I think the airport PA is playing the smooth jazz version of "smells like teen spirit"
[10:34:03] <linuxcnc-build> build #2035 of 4006.deb-lucid-rtai-i386 is complete: Failure [4failed shell_1] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/4006.deb-lucid-rtai-i386/builds/2035 blamelist: Sebastian Kuzminsky <seb@highlab.com>
[10:38:00] <seb_kuzminsky> what? that's not supposed to fail now
[10:38:27] <seb_kuzminsky> jepler: i loved that song, back when i had long hair
[10:42:21] <seb_kuzminsky> grr, it's another permissions problem in deb archive management
[10:42:29] <seb_kuzminsky> i thought i fixed that
[10:43:23] <jepler> pcw_home: are any of your pci(e) devices capable of delivering interrupts to the PC?
[10:43:42] <pcw_home> All are
[10:43:45] <jepler> the topic of interrupt-driven rtapi/hal threads was raised again at the txrx event
[10:44:24] <jepler> I think it's feasible in uspace to wait for an interrupt from a pci device
[10:45:14] <jepler> though there's some stuff to arrange before you could say "this is a nominal 1kHz thread, but it is driven by this interrupt instead of a linux clocksource"
[10:45:20] <pcw_home> Currently of the DPLL is installed it can generate interrupts
[10:45:35] <pcw_home> s/of/if/
[10:46:25] <jepler> right now my development time is directed at this user interface library development, but maybe sometime this winter I'll look at it, particularly if I can do it on the 5i20 that I already have
[10:46:55] <pcw_home> The original hostmot stuff did that (worked under DOS) clock was a submultiple of PWM rate
[10:47:01] <jepler> if I need a firmware with DPLL and I can't sort it out myself, I'll ping you when I get ready to actually work on it
[10:47:30] <jepler> >>> Posted May 3, 2007 20:06 UTC (Thu) by xav (subscriber, #18536) [Link]
[10:47:33] <jepler> People in the embedded space don't do prototypes. They hack something until it works, then it's done.
[10:47:34] <pcw_home> OK
[10:47:36] <jepler> <<<
[10:47:54] <jepler> (why am I reading years-old internet comments by jerks?)
[10:49:16] <pcw_home> Funny the DPLL is a way of getting along with LinuxCNC timing, but can also free run and be the master (set gain to 0)
[10:52:33] <pcw_home> Interrupt driven threads could also be used for Ethernet (but requires all possible read data every packet)
[10:52:34] <pcw_home> this would cut the packet traffic to 2 per cycle
[11:08:18] <decimad2> so apart from the kinematics stuff I just crushed into because of my dad's cnc... I'm designing a field oriented servo control with ptp synch... is there a more appropriate way than to present a "virtual stepper" with some delay for smoothing to the software?
[11:24:34] * cradek stops for breakfast before crossing the grits line
[11:31:26] <seb_kuzminsky> linuxcnc-build: force build --branch=2.7 --revision v2.7.0-pre2 0000.checkin
[11:31:27] <linuxcnc-build> build forced [ETA 1h08m16s]
[11:31:27] <linuxcnc-build> I'll give a shout when the build finishes
[11:33:15] <pcw_home> "some delay" if its a significant time will cause trouble with Linuxcnc real time inputs (homing,probing, threading, tapping etc)
[11:37:37] <cradek> ha, reinventing the buggy whip with christmas tree tinsel
[11:39:44] <pcw_home> I'm old enough to remember the metal stuff
[11:40:01] <pcw_home> wonder if it was tin
[11:40:18] <cradek> I wonder!
[11:49:32] <seb_kuzminsky> ok i think i found the cause of the permission problem
[11:50:04] <seb_kuzminsky> the buildmaster runs inside twistd, and twistd explicitly sets its umask to 077 on startup "for security reasons"
[11:50:30] <seb_kuzminsky> i think i can ask it to reset it to 022 through the config file, i'll try it after this build finishes
[12:28:21] <seb_kuzminsky> bummer, that didnt work
[12:31:43] <seb_kuzminsky> the done thing seems to be to modify the source code of twistd :-/
[13:12:20] <seb_kuzminsky> it's funny, the buildslave umask can be controlled, but not the buildmaster
[13:43:20] <dirty_d> the userspace hal interface is just basically shared memory and nothing more?
[13:44:54] <dirty_d> like when a pin signal changes linuxcnc just writes to the shared memory, but doesnt notify the driver that anything has changed?
[13:45:32] <dirty_d> and it just has to poll the state?
[13:50:28] <micges-dev> dirty_d: if you talk about halui, it's structured like any other gui, but instead of buttons and checkboxes it's controlled by hal pins
[13:51:53] <dirty_d> i dont think thats what im talking about, im talking about whats in i.e. gs2_vfd.c
[13:52:10] <dirty_d> a userspace vfd driver it seems
[13:52:57] <dirty_d> it looks like the hal pins are really just named variables in shared memory right?
[13:55:34] <micges-dev> I don't know this driver
[13:55:52] <dirty_d> well im really just curious about the userspace hal api
[13:56:11] <dirty_d> not specifically this driver
[13:56:25] <dirty_d> but it seems like there is no concept of an event really
[13:56:54] <micges-dev> nope
[13:56:56] <dirty_d> its up to the driver to poll the state of the shared hal memory
[13:58:02] <micges-dev> hal triggers were many times discussed but there is no clean design of it
[13:58:33] <dirty_d> wouldnt it make sense for a userspace driver to be a shared library that exports an abstract class where it calls a function with a given pin and level etc to notify it?
[13:58:39] <micges-dev> like changing hal value -> generate event
[13:59:41] <dirty_d> and if the driver happens to need to do other stuff it can create another thread when linuxcnc loads it
[14:16:38] <dirty_d> something like this http://codepad.org/l44QuChq
[14:17:44] <dirty_d> linuxcnc would load the library, get a pointer to the class instance, call getPins to see what pins the driver has, and setPinState to change the state and optionally set seconds so that it could i.e be sent to external hardware and be buffered
[14:17:49] <dirty_d> just using my imagination
[14:39:48] <decimad2> pcw_home what would you think about 3-5 pulses at max-speed?
[14:40:15] <PCW> pulses?
[14:41:29] <CaptHindsight> dirty_d: http://wiki.linuxcnc.org/cgi-bin/wiki.pl?EMC_Components have you seen this?
[14:42:38] <decimad2> the equivalent time of 3-5 stepper pulses at max-speed
[14:43:29] <decimad2> gotta reboot :(
[14:48:43] <CaptHindsight> dirty_d: an updated diagram with HAL is here http://linuxcnc.org/docs/html/code/Code_Notes.html
[14:59:01] <dirty_d> no, but I just looked at it
[14:59:10] <PCW> 100 or 1000 stepper pulses is fine (that is a 1 MHz step rate is fine at a 1 mS servo thread)
[14:59:12] <PCW> if you have extreme acceleration or high accuracy requirements you may need a faster servo thread
[15:08:30] <PCW> My current stepgen config for testing jitter tolerance moves at 40 IPS (~60 M/minute)
[15:08:32] <PCW> and has ~1.25 G acceleration. it keeps within about 30 uinches with ~100 usec thread jitter
[15:08:34] <PCW> This is closed loop PID controlled at a 2 KHz servo thread
[15:08:35] <PCW> At higher accelerations the step generator will probably need incorporate acceleration in the rate generator
[15:08:37] <PCW> to avoid velocity steps (for example in my current setup it takes about 80 ms to get from 0 to 40 IPS so
[15:08:38] <PCW> only 160 velocity steps total at 2 KHz
[15:13:55] <micges-dev> PCW: if stepgen will have accel, how fast you can update steprate?
[15:15:22] <PCW> probably some fraction clocklow to keep adder sizes within reason, but say 1 MHz or so
[15:15:30] <PCW> of clocklow
[15:17:20] <PCW> I would probably use a rate multiplier to allow scaling of the acceleration number (so the ~ 1 MHz would be variable to set accel scaling)
[15:19:46] <PCW> but in any case, much faster than the servo thread
[15:21:00] <PCW> so now you update the steprate and the accel at the servo thread rate
[15:25:02] <micges-dev> I see
[15:32:19] <jepler> home!
[15:43:55] <JT-Shop> yea! how did it go?
[15:51:09] <seb_kuzminsky> JT-Shop: the hackfest was great fun, the hosts were interesting friendly people, and we even did some useful work in between the drinking and schmoozing
[15:53:24] <JT-Shop> lol sounds like fun, wish I could have gone
[15:58:08] <seb_kuzminsky> i hope you can make it to the hackfest next year
[15:58:34] <seb_kuzminsky> oh, we did one thing that i wanted to ask you about
[15:58:55] <seb_kuzminsky> jmk wanted the hal docs reorganized a bit, i hope you approve
[16:00:25] <seb_kuzminsky> http://git.linuxcnc.org/gitweb?p=linuxcnc.git;a=commit;h=51eaa0ab8533c3f0a8b041c676c8e297640dfd53
[16:00:54] <seb_kuzminsky> starting with "docs: don't call the HAL tutorial Advanced" and ending with "docs: minor reorg of HAL docs"
[17:03:24] <jepler> jthornton: it was a good event, and seb and I bit off a big project to work on together
[17:13:58] <jepler> > It took me a while to understand what was going on, but after thoroughly studying it I developed a rigorous explanation for this staggering decline in standards: kids today.
[17:16:24] <decimad2> which decline in standards?
[17:17:19] <kwallace1> Crud, my Shizuoka mill just blew a stepper drive. It is a unipolar 45V 8 amp driving a NEMA42 motor. I noticed the Gecko G-203V runs up to 7 amps and 80V. Has anyone run a NEMA42 stepper with a Gecko?
[17:17:49] <jepler> decimad2: sorry, just quoting random web pages into the channel
[17:18:03] <jepler> decimad2: http://www.slate.com/blogs/lexicon_valley/2014/10/21/single_quotes_or_double_quotes_it_s_really_quite_simple.html
[17:19:06] <jepler> kwallace1: crud :-/ no personal experience with steppers that big, sorry
[17:19:24] <decimad2> It just so happens that I can quotations in any way and don't really care... :)
[17:19:33] <decimad2> add a "read" in there
[17:24:10] <kwallace1> Another option is to pull the driver and try to fix it, but of course, I'm in the middle of milling an overdue part. BTW, here is the drive: http://www.wallacecompany.com/machine_shop/Shizuoka/00001-1a.jpg
[17:31:52] <seb_kuzminsky> bummer kwallace1! seems like equipment is most likely to fail when you need it most
[17:33:02] <seb_kuzminsky> ok, 2.7.0~pre2 built and made debs, i'll put them on wlo tonight(?) and send out some kind of announcement
[17:33:07] <seb_kuzminsky> rob ellenberg will be happy
[17:59:54] <kwallace1> I'm not too upset, the drive is 30 years old.
[18:03:17] <ssi> kwallace1: do you know how much current the motor wants?
[18:08:45] <kwallace1> ssi, I seem to recall the drive is rated at 8 Amps. The output fuses are 12 Amp. Schematic here: http://www.wallacecompany.com/machine_shop/Shizuoka/Driver%20Schematic1.jpg
[18:10:41] <ssi> yeah, but I'm more curious about the motor :)
[18:11:08] <ssi> also, is the motor strictly unipolar, or can it be wired bipolar?
[18:17:00] <PCW> looks simple enough to fix, what blew?
[18:23:05] <PCW> the motor is 6 wire so can be run Bipolar
[18:35:19] <kwallace1> Yeah, I can do bipolar. The stepper motors can too. The motors don't have any markings so I don't know the specs.
[18:37:05] <ssi> markings would be too convenient :)
[18:37:54] <kwallace1> I think I found a bad transistor, which is a TIP36C : http://www.digikey.com/product-detail/en/TIP36C/497-2604-5-ND/603629
[18:55:22] <seb_kuzminsky> jepler: am i on the right track here?
[18:55:28] <seb_kuzminsky> i want to fix this: http://buildbot.linuxcnc.org/buildbot/builders/1404.rip-wheezy-rtpreempt-amd64/builds/892/steps/compile/logs/warnings%20%2882%29
[18:55:34] <seb_kuzminsky> by doing this: http://pastie.org/9669234
[20:23:57] <jepler> seb_kuzminsky: looking ..
[20:24:31] <jepler> seb_kuzminsky: looks plausible
[20:29:09] <jepler> > The Parallel Port's interrupt request is not used for printing under DOS or Windows. Early versions of OS-2 used them, but don't anymore
[21:04:55] <cradek> the more I looked at (for) bug 397 the more units weirdness I saw in AXIS
[21:05:08] <cradek> seems there are several things to fix
[21:05:30] <cradek> I got half of them half-fixed last night
[22:00:07] <skunkworks_> wow - 2.7 is about to be released?
[22:00:19] <cradek> no - ready to start stabilizing
[22:00:48] <cradek> skunkworks_: got the microscope-camera today - it is as neat as you said
[22:01:59] <skunkworks_> great - I am glad you like it.
[22:02:14] <skunkworks_> I only got it because someone on the list mentioned it
[22:02:20] <skunkworks_> and it was cheap :)
[22:03:42] <CaptHindsight> http://imagebin.ca/v/1c9zuonpOcyi 100um gap between the jaws of calipers with that scope
[22:04:02] <skunkworks_> I think it was bari... :)
[22:04:13] <CaptHindsight> http://imagebin.ca/v/1cA0LyuHgway better shot
[22:06:09] <skunkworks_> logger[psha]:
[22:06:10] <logger[psha]> skunkworks_: Log stored at http://psha.org.ru/irc/%23linuxcnc-devel/2014-10-23.html
[22:07:26] <CaptHindsight> it's ok for ~$20