#emc-devel | Logs for 2009-07-08

[17:25:05] <cradek> micges: http://www.linuxcnc.org/component/option,com_kunena/Itemid,20/func,view/catid,20/id,438/limit,6/limitstart,6/lang,en/#495
[17:25:47] <cradek> micges: maybe you are interested in this - unfortunately it sounds like it's in the least convenient form possible!
[17:26:33] <jepler> ugh, indeed it is
[17:29:57] <micges> heh
[17:30:25] <micges> where did he mentioned earler about bugs on it?
[17:30:47] <cradek> I don't know, sorry
[17:31:32] <micges> ok
[17:33:07] <jepler> if you are logged into linuxcnc.org, you can see his other posts by clicking on his name in that post, then clicking on "forum".
[17:33:26] <micges> I see it
[17:33:41] <jepler> oh, or click "<< start" to go back to the beginning of the thread..
[17:33:42] <cradek> above his message you can go to page 1, which is earlier in the thread
[17:34:45] <micges> did you checked his program?
[17:35:08] <cradek> no, I didn't do anything
[17:38:15] <micges> indeed his program beahave strange on master, I
[17:38:20] <micges> 'I'll check it
[17:38:43] <cradek> great, thanks
[17:54:42] <micges> ok, his bug is repetable, bug is in readahead code, last time I touched readahead it hangs task ;)
[17:55:03] <micges> I'll debug some info of that later
[17:56:00] <micges> I saw 2 more bugs running this program
[17:56:14] <micges> 1. unknown task state after motion queue full
[17:57:03] <micges> 2. no readahead limit of (default) 1000 lines, task try to load program whole at once
[17:57:16] <micges> bbl
[19:07:51] <CIA-1> EMC: 03jepler 07master * rbc39385f738b 10/.gitignore: the halscope autosave file should probably never be in git
[19:23:59] <micges> hmm after clone I have less *.hal files in configs than after make clean, why?
[19:24:08] <micges> are there some generated files ?
[19:24:50] <jepler> some are copied from configs/common
[19:25:04] <jepler> they probably aren't cleaned by make clean because nobody bothered to write that line of makefile
[19:25:30] <micges> jepler: where to look?
[19:28:08] <jepler> for rules executed when you 'make clean', search for "^clean:" in the top-level Makefile
[19:28:30] <jepler> for files copied in configs directory, look near COPY_CONFIGS
[19:28:34] <jepler> also in the top-level Makefile
[19:28:48] <micges> that Makefile isn't generated?
[19:29:09] <jepler> no, Makefile and the Submakefiles are written by humans
[19:29:15] <jepler> Makefile.inc is generated from Makefile.inc.in
[19:29:37] <micges> I see
[19:29:46] <jepler> we don't use GNU automake or other makefile-making syst
[19:29:51] <jepler> ms
[19:29:53] <jepler> syste
[19:29:56] <jepler> systems
[19:30:01] <jepler> * jepler kicks his keyboard
[19:30:06] <micges> heh
[19:30:20] <jepler> hi seb
[19:30:26] <seb_kuzminsky> hi there
[19:30:36] <seb_kuzminsky> how's your servo zenbot?
[19:30:59] <jepler> the first axis seems nice
[19:31:06] <jepler> tonight we'll finish the other two mounts
[19:31:11] <seb_kuzminsky> cool :-)
[19:31:26] <jepler> I also have to build the other two connectors to the servo amp, and then tune the whole thing
[19:31:31] <jepler> maybe this weekend I can mill something
[19:32:35] <jepler> I was reading recently about physical random number generators based on avalanche noise .. I might use that as a project
[19:33:17] <cradek> http://timeguy.com/cradek-files/emc/IMAG0045.jpg
[19:33:38] <cradek> this is the Z mount
[19:33:54] <seb_kuzminsky> nice finish in that big bored hole
[19:33:58] <jepler> oooh shiny bore
[19:34:04] <cradek> yep 0.749
[19:34:29] <cradek> that's the critical dimension - worked great
[19:35:00] <seb_kuzminsky> cradek: are you doing rigid tapping?
[19:35:11] <jepler> (similar to this: http://www.postcogito.org/Kiko/T12RandomNumberGenerator.html )
[19:35:14] <cradek> the lathe will do it, the mill is not running emc
[19:35:37] <cradek> I have used it only a few times on the lathe
[19:36:59] <cradek> I hope to do some form tapping with it soon, for a little project (maybe a couple dozen parts)
[19:37:05] <jepler> (hm, serial-port powered?)
[19:37:28] <seb_kuzminsky> you can pull quite a bit of power off the cts pin
[19:38:20] <seb_kuzminsky> uh rts i guess
[19:39:01] <cradek> traditionally, all pins have the same drivers
[19:39:13] <cradek> and yeah they are fairly high current - serial cables were long.
[19:39:18] <seb_kuzminsky> cts is input to the serial port, rts is output from the serial port
[19:39:42] <jepler> cradek: you might be confusing hardware made in the 90s with hardware made today
[19:39:50] <cradek> hm
[19:40:03] <jepler> you might think hardware made today with those "rs232" ports complies with rs232, but I wouldn't take that bet
[19:40:13] <seb_kuzminsky> i guess these days folks just use usb
[19:40:48] <jepler> the text says you need at least 12V for the avalanche effect to work .. I don't understand it, so I'd just have to take it at face value
[19:40:59] <seb_kuzminsky> cradek: is your mill manual?
[19:41:14] <jepler> (the other version of this project I've found uses a 12V wall-wart)
[19:41:14] <cradek> no, it's an ancient cnc
[19:41:16] <seb_kuzminsky> or running some other controller? mach3 mauye? ;-)
[19:41:24] <cradek> boss8
[19:41:28] <cradek> 8 >> 3
[19:41:44] <seb_kuzminsky> 8 >> 3 == 1 ;-)
[19:41:59] <cradek> 8 ยป 3
[19:42:19] <cradek> 8 is much greater than 3
[19:42:20] <seb_kuzminsky> so the arcs on the z mount were done via some conversational interface?
[19:42:24] <seb_kuzminsky> or mdi?
[19:42:46] <cradek> we laid out the arcs in autocad, and then generated the gcode with REALIZE
[19:43:05] <seb_kuzminsky> oh ok cool
[19:43:05] <cradek> the whole thing was one program - 7 tools
[19:43:19] <cradek> (even the tapped holes)
[19:43:56] <seb_kuzminsky> the rectangular outside of the workpiece was done manually before running the program, and the workpiece held in the vise?
[19:44:11] <jepler> no, we made a fixture
[19:44:13] <cradek> nope, it was screwed to a fixture plate, and the outside was cut out
[19:44:22] <cradek> I have a photo... one minute.
[19:44:42] <jepler> the 4th side was cut in the vise later, though, because the second mounting hole is outside the plate
[19:44:58] <jepler> the first mounting hole is in the center of the non-tapped holes
[19:45:15] <jepler> iirc cradek did that as a manual operation
[19:45:17] <jepler> measure, cut twice
[19:45:41] <seb_kuzminsky> "i've cut it twice already, and it's still too small"
[19:46:49] <cradek> http://timeguy.com/cradek-files/emc/IMAG0041.jpg
[19:47:00] <cradek> that's the last cutting operation with a tiny 1/8" end mill
[19:47:22] <seb_kuzminsky> neat
[19:47:37] <cradek> you can barely see the two shoulder bolts holding it to the triangle scrap fixture
[19:47:56] <jepler> you can see one, anyway. the other's pretty buried in swarf.
[19:48:06] <cradek> the fixture has holes drilled through below the tapped holes, so a spiral point tap could be used in the tapping head.
[19:48:07] <seb_kuzminsky> second-to-last operation, before freeing the part from the workpiece?
[19:48:19] <cradek> yes
[19:48:47] <seb_kuzminsky> that's really cool you guys
[19:49:23] <cradek> very successful: 3/3 usable parts, no broken things :-)
[19:49:41] <jepler> I'm wary of declaring victory quite yet
[19:50:12] <jepler> there are still 5 press fits and 2 fiddly little connectors to be made
[19:50:26] <cradek> what are the fiddly connectors?
[19:50:37] <jepler> the encoder & motor to mesa connectors
[19:50:47] <cradek> ah, that's nmfp!
[19:50:53] <jepler> nope
[19:51:13] <jepler> but those other parts you were alluding to earlier are, I need them to set up my electronics box nicely
[19:51:25] <seb_kuzminsky> jepler: are you planning to mill a little copper-clad board to solder the connectors to, like those you showed me at Fest?
[19:51:54] <jepler> seb_kuzminsky: not sure. I haven't seen any indication that the bouncing signals are a problem when connected to a mesa board..
[19:52:08] <jepler> a little bounce would just make the pwm funky, not cause extra or lost steps
[19:52:13] <cradek> ah, well that's not done yet... it will be fun. should be able to make about 1 per minute once it's going I think.
[19:53:33] <jepler> yeah it seems like it should go pretty fast
[19:55:08] <jepler> turn, drill, tap, part. then in a second mounting, face
[19:55:38] <cradek> too bad about the second mounting, but I don't see a way around it.
[19:55:41] <jepler> half that time will be waiting for toolchanges and spindle-at-speed
[19:55:58] <jepler> we talked about having the hole go all the way through, but I don't like that
[19:56:06] <cradek> nope, that would be bad.
[19:56:56] <cradek> two drills - one will be a spot drill so the hole is centered and chamfered for the form tap
[20:01:04] <CIA-1> EMC: 03micges 07joints_axes3 * r73b179e7efa0 10/configs/ (72 files in 36 dirs): change sample configs to follow motion pin names change axis.nn => joint.nn
[20:01:17] <cradek> wow
[20:01:45] <jepler> build a fixture that lets you put about 25 pieces on it, then do the facing operation on the mill instead.
[20:02:20] <cradek> not sure how you'd clamp a round thing on end...
[20:02:40] <jepler> this is probably not a good idea, really
[20:02:49] <jepler> it would be posts that give a "just snug enough" fit against the threads
[20:03:00] <jepler> then you just touch it with a mill bigger than the outer diameter
[20:03:25] <cradek> half the cut would tend to unscrew it
[20:03:55] <jepler> just a round post, not a threaded one
[20:04:08] <cradek> ok, that's worse :-)
[20:04:16] <jepler> yeah it's a terrible idea
[20:04:27] <jepler> with a hole for vacuum at the peak of each post
[20:04:44] <seb_kuzminsky> is the material Al?
[20:04:50] <cradek> or brass
[20:05:16] <jepler> is there some way to support it in the lathe, so that parting doesn't leave the dot?
[20:05:27] <cradek> nope
[20:05:41] <jepler> (I'm sure you would have thought of any non-nutty scheme)
[20:06:07] <seb_kuzminsky> what does this part look like?
[20:06:32] <cradek> tapped hole on one end, flat foot on the other end
[20:07:00] <seb_kuzminsky> like a flat-head screw, except with female threads?
[20:07:16] <cradek> yeah, something like that
[20:07:55] <seb_kuzminsky> and you want to face the foot, where it was parted off from the bar stock?
[20:08:00] <cradek> yes
[20:08:03] <jepler> right
[20:09:23] <jepler> something like this: http://emergent.unpy.net/files/sandbox/nothing.png
[20:09:41] <seb_kuzminsky> right
[20:09:54] <cradek> the top should be the diameter of the matching SHCS
[20:10:12] <cradek> also, I'd use right-hand threads instead of left-hand
[20:10:19] <jepler> well that would be very practical
[20:11:44] <cradek> brb
[21:00:57] <alex_joni> hmm.. I bet the "fix" for emctaskmain.cc breaks step/pause
[21:02:43] <alex_joni> might have been this that broke it: http://git.linuxcnc.org/gitweb?p=emc2.git;a=commitdiff;h=78a8f493bed278f1705e47ade0475d803cc08785
[21:03:24] <jepler> worth a try 'git revert'ing that patch and testing again
[21:04:07] <alex_joni> well.. I've never seen the "bug", cause I don't have such long programs
[21:04:20] <alex_joni> and it's bedtime too ;)
[21:07:00] <jepler> how long a program is required?
[21:07:11] <alex_joni> I think > 50k lines or so
[21:07:43] <alex_joni> http://www.linuxcnc.org/component/option,com_kunena/Itemid,20/func,view/id,438/catid,20/limit,6/limitstart,0/lang,en/
[21:07:47] <alex_joni> there is one attached on that page
[21:09:12] <micges> Isn't that importand that all g codes in that file are G1 ?
[21:10:09] <alex_joni> http://dsplabs.upt.ro/~juve/emc/bobo.nc
[21:10:49] <alex_joni> micges: I think all G1's can lead to motion queue to fill up faster
[21:11:05] <alex_joni> good night all
[21:11:17] <micges> good night alex
[21:15:29] <jepler> the how to reproduce is what .. select "step" then "continue"?
[21:16:35] <micges> load, step, resume will show bug
[21:17:21] <jepler> and what will it do?
[21:17:49] <micges> run 1000 and stop
[21:17:56] <micges> up to 100
[21:18:03] <micges> 1000*
[21:18:10] <cradek> wonder if this was kirk wallace's problem with that encoder disk thingy
[21:18:22] <cradek> maybe it just happened sometimes because he used pause differently
[21:19:18] <jepler> reproduced here as well
[21:19:26] <jepler> doesn't need a 50kline program either
[21:20:26] <jepler> I ran the output of this: http://emergent.unpy.net/files/sandbox/makebig.py
[21:23:53] <jepler> (can be directly loaded by sim/axis)
[21:31:29] <jepler> alex is right that reverting 78a8f493 makes the behavior go away, but that's because the sequence T S (step resume) doesn't work.. the closest sequence that works is T P S
[21:32:57] <jepler> with master, the sequence T P S doesn't exhibit the bug
[21:33:13] <jepler> so what changes in task state by pressing P?
[21:34:47] <jepler> -interp_state reading
[21:34:47] <jepler> +interp_state paused
[22:34:58] <micges> good night all
[22:35:39] <CIA-1> EMC: 03micges 07joints_axes3 * rb16df2580035 10/src/emc/usr_intf/axis/scripts/axis.py: update joint pins names