#linuxcnc-devel | Logs for 2014-12-06

[10:26:50] <seb_kuzminsky> ok here's one i feel good about
[10:26:50] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 2a3699f 06linuxcnc 10docs/src/Submakefile 10src/Makefile 10src/hal/components/Submakefile docs: reorg how manpages are generated * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=2a3699f
[10:26:50] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 0392ca1 06linuxcnc 10docs/src/Submakefile docs: switch to recursive expansion for manpage variables * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=0392ca1
[10:26:50] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 633e0ed 06linuxcnc 10docs/src/Submakefile docs: include linuxcnc.1 in the manpage index and pdf * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=633e0ed
[10:26:53] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 171de4e 06linuxcnc 10src/hal/user_comps/wj200_vfd/Submakefile docs: include the wj200 manpage in the html & pdf docs * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=171de4e
[10:26:57] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 5122891 06linuxcnc 10docs/src/Submakefile docs: build index.incl cleaner & more completely * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=5122891
[10:27:01] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 6116005 06linuxcnc 10docs/src/Submakefile docs: verify all manpages are in the html manpage index * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=6116005
[10:27:05] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 4c1ee25 06linuxcnc 10docs/src/Submakefile docs: select the man1 and man3 manpages correctly * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=4c1ee25
[10:27:09] <KGB-linuxcnc> 03Sebastian Kuzminsky 05seb/2.6/wj200-manpage-alt 02b2bb3 06linuxcnc 10(7 files in 5 dirs) docs: move the skeleton manpages out of the way * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=02b2bb3
[10:28:27] <seb_kuzminsky> if there are no objections i'll push that change to the 2.6 docs build system
[10:28:46] <seb_kuzminsky> it fixes many omissions of manpages in the html manpage index
[10:28:48] <seb_kuzminsky> bbl
[11:48:37] <jepler> seb_kuzminsky: read the diffs, didn't spot any problems
[11:51:42] <kwallace> Let's say I want to cut a G2 helix for a 5/8-11 thread .4" long. That's 4.4 turns, so I use P = 4 and calculate were the arc end should be and enter that for x and y. I expect the 4 turns to be cut then the arc, but noooo, I get three turns and the arc. Where am I going wrong?
[11:53:09] <kwallace> G2 X -0.6702 Y -0.4841 Z -0.3880 I -0.8267 J 0.0000 P 4.0000
[11:54:55] <kwallace> ... Oops, that Z doesn't look right.
[11:56:52] <KimK> Yes, G2's Z has to be 1/11, and you repeat the G2 as many times as you need.
[11:58:31] <seb_kuzminsky> jepler: thanks
[11:58:54] <kwallace> My line above is correct for that particular move, here is a better command set.
[11:58:57] <kwallace> G0 Z 0.0000
[11:58:57] <kwallace> G1 X 0.8060
[11:58:57] <kwallace> F 16.82 (Arc Feed, inches/minute)
[11:58:57] <kwallace> G2 X -0.6534 Y -0.4720 Z -0.4000 I -0.8060 J 0.0000 P 4.0000
[11:59:33] <cradek> kwallace: P is "number of full or partial turns" so to get 4.4 turns you will have to specify P5
[11:59:49] <cradek> > For example, if a 180 degree arc is programmed with a P2, the resulting motion will be 1 1/2 rotations. For each P increment above 1 an extra full circle is added to the programmed arc.
[12:00:39] <cradek> I think the docs spell it out correctly
[12:00:45] <cradek> http://linuxcnc.org/docs/html/gcode/gcode.html#sec:G2-G3-Arc
[12:02:05] <kwallace> But that is not intuitive and when I read the doc I tried to make the doc fit my understanding.
[12:02:53] <kwallace> I would think that if one were to make a simple arc, P would be 0.
[12:04:13] <cradek> if you want exactly 1 or exactly 2 full turns what would your intuition tell you to use?
[12:04:48] <cradek> wait, I don't want to argue about this, I only care that the docs correctly describe how linuxcnc works, and I think that is the case
[12:05:24] <cradek> P being an integer, but allowing nonintegral turns is bound to be nonintuitive in some situations
[12:05:53] <kwallace> Yes agreed, I just want to set my script to get the correct outcome, so I'll add one to my P and see how it goes.
[12:07:03] <cradek> yes P1 (the default) gives you up to exactly one full turn, P2 gives you up to 2 full turns
[12:11:43] <kwallace> On the other side, we want the operator to be able to read the g-code and tweak it. If they see P = 5 plus an arc, my intuition would be that getting four turns plus an arc would be wrong even though, in this case makes a good part.
[12:13:24] <kwallace> Anyway, I thought my math was wrong, but it looks like my P foo is, and I know how to deal with it. Thank you.
[12:14:47] <kwallace> Or could do as KimK suggests and just program the turns and arc separately -- not much ambiguity there.
[12:17:49] <KimK> Do you *have* to have 4.4 turns? Might make life a lot easier if you start high and take 5 (or 6) turns.
[12:21:23] <kwallace> This is the application: http://wallacecompany.com/tmp/Screenshot_tmill.png
[12:23:07] <kwallace> I set however many turns that fit in the Z range.
[12:23:52] <kwallace> Another curious thing, divmod and modf give me different results, but I haven't taken the time to investigate.
[12:24:01] <kwallace> --kw helix revs = 4.400440044
[12:24:01] <kwallace> --kw divmod = (4.0, 0.036400000000000043)
[12:24:01] <kwallace> --kw mod_revs = (0.4004400440044007, 4.0) full_revs = 4.0
[12:24:01] <kwallace> --kw arc degrees = 144.158415842 rad = 2.51603900089
[12:26:45] <KimK> Looks great. Well, if it was me (I like to avoid headaches), I'd specify an integer number of turns (5?), and the lie to it about z start and say it's 5/11 above z end. But that's just me, many ways to skin the proverbial cat. Is that your app pictured there?
[12:27:18] <KimK> s/and the lie/and then lie/
[12:29:28] <KGB-linuxcnc> 03Sebastian Kuzminsky 052.6 02b2bb3 06linuxcnc fast forward * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=02b2bb3
[12:29:42] <KGB-linuxcnc> 05seb/2.6/wj200-manpage d2cd0f9 06linuxcnc 04. branch deleted * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=d2cd0f9
[12:29:47] <KGB-linuxcnc> 05seb/2.6/wj200-manpage-alt 02b2bb3 06linuxcnc 04. branch deleted * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=02b2bb3
[12:31:25] <seb_kuzminsky> does anyone know of any blockers for 2.6.5?
[12:34:28] <kwallace> KimK, my app resembles this one: https://www.youtube.com/watch?v=WizzhG2JrjQ at about :22
[12:36:56] <pcw_home> what in 2.6.5? wasn't 2.6.4 just released a couple weeks ago?
[12:40:20] <seb_kuzminsky> 2.6.4 was released on Nov 1
[12:41:23] <seb_kuzminsky> since then we've had some fixes to halcompile, a fix for a dropped mdi bug in task, a fix to gmoccapy, and improved docs
[12:45:22] <pcw_home> OK
[12:45:40] <pcw_home> I noticed a minor bug on latency-histogram ( it doesn't clean up properly if you abort it with control C)
[12:48:34] <seb_kuzminsky> whoops!
[12:49:23] <pcw_home> the gui X works
[12:50:17] <seb_kuzminsky> latency-histogram was written by dgarr
[12:50:28] <seb_kuzminsky> normally i'd ask you to poke him about this bug report
[12:50:46] <seb_kuzminsky> but he's deep in his moveoff work, and i dont want to distract him from that
[12:52:23] <seb_kuzminsky> i'll take a look at it, though i suck at tcl
[12:52:59] <pcw_home> Sure, not a big deal, i just get snagged on it occasionally
[12:53:12] <seb_kuzminsky> it's a papercut-class bug
[12:54:00] <pcw_home> :-)
[12:54:33] <pcw_home> Pretty decent latency for Preemt-RT:
[12:54:34] <pcw_home> http://imagebin.ca/v/1jjIAesLlzHW
[12:55:13] <seb_kuzminsky> sweet!
[12:55:21] <seb_kuzminsky> that's when the new cpu you got in the mail, with more cache?
[12:55:35] <kwallace> That's better, so far: http://wallacecompany.com/tmp/Screenshot_tmill2.png
[12:56:16] <pcw_home> Yes $15.00 e8500
[12:58:25] <KimK> Hi Seb, are the recent changes to comp described in the 2.6 standard (preinstalled) docs? If not, where should I go to read up?
[13:00:38] <seb_kuzminsky> KimK: jepler added extra_link_args, documented here: http://www.linuxcnc.org/docs/2.6/html/hal/comp.html#_options
[13:00:55] <seb_kuzminsky> bbl
[13:01:05] <KimK> OK, thanks
[15:18:13] <memfrob> i dont know why but preempt-rt latency is always around 100us for me and i've tried all different kinds of kernel configs too and it never dropped
[15:19:29] <memfrob> i've used the tracing software and it was never helpful, something about cpu idle and changing all the cpu idle state stuff in the kernel never made a bit difference nor idle=poll
[15:19:39] <pcw_home> It seems cache size is important
[15:19:52] <memfrob> L2 or L3?
[15:20:25] <pcw_home> probably either
[15:21:30] <pcw_home> The CPUs I have that have the best Preemt-RT performance have the larges caches (one L2 and the other L3)
[15:21:32] <pcw_home> still just a guess
[15:22:11] <pcw_home> higher thread rates improve latency (less likely to have RT stuff dumped maybe)
[15:23:08] <memfrob> AMD abandonded L3 cache it's all L2 now, mine says it only has 4MB L2 without L1 or L3 caches
[15:26:25] <pcw_home> The best performing CPUs for Preemt-RT I have are a E8500 and a G3258
[15:26:26] <pcw_home> The E8500 is a 3.16 GHz Core Duo with 6MB L2 cache, the G3258 is a 3.2 GHz Haswell with 3MB L3 cache
[15:27:08] <memfrob> ah so you get good numbers with intel, i see.
[15:27:12] <pcw_home> the E8500 has better latency than the G3258 even though the G3258 is quite a bit faster
[15:27:16] <JT-Shop> the description of G38 is not clear, do the variables hold the position including any offsets or the machine position after a successful prob?
[15:27:20] <JT-Shop> e?
[15:27:52] <pcw_home> I dont have any current AMD CPUs (well except low power fanless ones)
[16:11:11] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/moveoff c0fa4f6 06linuxcnc 10scripts/moveoff_gui moveoff_gui: track enable for external setters * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c0fa4f6
[16:11:11] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/moveoff c2bcd71 06linuxcnc 10(9 files in 3 dirs) moveoff_gui: options for autoresume with delay * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c2bcd71
[17:29:28] <pcw_home> Hmm I think the residual untuneable errors in the stepgen are because FF2 is 1 cycle late
[17:29:30] <pcw_home> which is why the TP should provide _forward_ accel and velocity, and the PID not make its own with differences
[17:47:50] <pcw_home> of course maybe I shouldn't be worrying about .00002" errors at 20IPS/1.2G accell
[19:22:36] <skunkworks> heh
[19:29:30] <skunkworks> comutator is screwed up on the x lathe servo. Bulge in it.
[19:29:36] <skunkworks> (why i broke the brushes)
[19:31:01] <pcw_home> Thats bad
[19:31:30] <Tom_itx> wonder what caused the bulge
[21:52:27] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/moveoff d71977c 06linuxcnc 10scripts/moveoff_gui 10src/hal/components/moveoff.comp moveoff: typos, some cleanup * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=d71977c
[21:57:59] <dgarr> http://youtu.be/KY6hx7WBkO8
[22:07:47] <Tom_itx> is that enabled by default or do you add it in the ini or something?
[22:07:57] <Tom_itx> good stuff btw...
[22:10:55] <dgarr> http://git.linuxcnc.org/gitweb?p=linuxcnc.git;a=tree;f=configs/sim/axis/moveoff;h=68eab7343317c233faee7ee276982527bb173aa8;hb=refs/heads/dgarr/moveoff
[22:11:38] <Tom_itx> will that come in a git pull from master?
[22:13:24] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/moveoff e2388c6 06linuxcnc 10configs/sim/axis/moveoff/README 10scripts/moveoff_gui 10src/hal/components/moveoff.comp moveoff: typos, some cleanup * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=e2388c6
[22:16:26] <dgarr> yes, it is based on master currently so, for example: git fetch; git branch --track tst origin/dgarr/moveoff; git checkout tst
[22:17:12] <Tom_itx> can't wait to get my cnc fixed to try it..
[22:28:33] <skunkworks> dgarr: Nice work! thanks for working on this.
[22:28:54] <dgarr> np