#emc-devel | Logs for 2008-08-22

[00:00:01] <SWPadnos> whatever one is needed for backwards compatibility with 6.06
[00:00:02] <skunkworks> pminmo: now that you have been running emc for a while - what do you think of it? Did you ever run mach?
[00:00:11] <BigJohnT> he is not going to directly edit the lyx files
[00:00:29] <pminmo> ./config is running now
[00:00:38] <BigJohnT> SWEET!
[00:00:49] <pminmo> emc2 I like it, a lot! linux, not as much
[00:00:58] <BigJohnT> I'm real close to Arkansas :)
[00:01:17] <SWPadnos> oh - I thought you had it building minutes ago ;)
[00:01:24] <skunkworks> pminmo: sorry I didn't get to meet you at the cncfest.
[00:01:25] <SWPadnos> I guess it was installing the build deps ;)
[00:01:40] <pminmo> yes I've run mach3 for years, emc2 is blowing it away in terms of rapids with steppers
[00:01:53] <pminmo> where you there?
[00:02:28] <skunkworks> yes - thursday thru sat.
[00:02:41] <skunkworks> or was it wednesday? I don't remember
[00:02:51] <pminmo> I was there during the day Thurs, left about 5pm
[00:02:59] <pminmo> to come home
[00:03:26] <pminmo> was there all day tues, wed
[00:03:40] <skunkworks> maybe next year :)
[00:04:03] <BigJohnT> pminmo: I smell some meat cooking talk you guys later
[00:04:08] <pminmo> rayh had me mesmorized for 3 days
[00:04:20] <BigJohnT> send me an email on how you make out
[00:04:41] <BigJohnT> thanks for the help SWPadnos
[00:04:49] <SWPadnos> sure
[00:05:38] <pminmo> will there be a fest next year?
[00:05:45] <SWPadnos> should be
[00:06:09] <SWPadnos> I think Roland has one more CNC workshop planned for his place. after than, who knows
[00:06:15] <SWPadnos> s/than/that/
[00:06:30] <pminmo> excellent
[00:08:38] <skunkworks> pminmo: I am not a linux person either.. It grows on you.
[00:09:35] <pminmo> ok my make finished with a bunch of errors because of not able to convert graphics files
[00:10:35] <SWPadnos> right - that would be the problem when you don't have the extra doc-building dependencies
[00:10:54] <SWPadnos> which someone who knows better will have to list or point you toward
[00:14:04] <pminmo> looks like the only pdf that was successful was the hal_user_manual
[00:14:33] <SWPadnos> probably no diagrams to convert in that one
[00:26:28] <pminmo> installed imagemagick and I think I'm good now
[00:26:48] <pminmo> i've got pdf docs with todays date
[00:27:01] <SWPadnos> nice
[00:31:37] <pminmo> later, thanks going to see if I have enough space to build a 6.06 setup for lyx
[00:32:11] <SWPadnos> you don't need to do that :)
[00:32:34] <SWPadnos> all you need is the correct version of kyx, which someone else knows (a) what it is and (b) how to install it
[00:32:38] <SWPadnos> lyx, that is
[00:32:43] <pminmo> ok, i was under the impression that my 8.04 setup wouldn't work
[00:33:24] <SWPadnos> I think there's a way - you may have to unstall a version of lyx from source or something, but you're an old pro at that now :)
[00:33:28] <SWPadnos> gah
[00:33:49] <pminmo> i'm up to huggies............
[00:33:59] <pminmo> or at least pullups
[00:34:10] <SWPadnos> heh
[00:35:11] <jmkasunich_> jepler: around?
[00:35:28] <jmkasunich_> I'm getting tracebacks from hal_input
[00:35:59] <jmkasunich_> halcmd: loadusr hal_input "Vendor=046d Product=c214"
[00:36:17] <jmkasunich_> spew snipped, then
[00:36:20] <jmkasunich_> raise LookupError, "No input device matching %r was found" % pattern
[00:36:20] <jmkasunich_> LookupError: No input device matching '"Vendor=046d Product=c214"' was found
[00:36:51] <SWPadnos> hmm
[00:37:12] <jmkasunich_> I also tried using "Attack 3" and Attack
[00:38:50] <SWPadnos> I think it's a permissions issue
[00:38:57] <SWPadnos> you need to adda udev rule for input devices
[00:39:05] <SWPadnos> I think it's at least partly explained in the manpage
[00:39:08] <jmkasunich_> yeah, I'm starting to think that too
[00:39:11] <jmkasunich_> right, partly
[00:39:38] <jmkasunich_> my rules.d dir doesn't have a file that sounds like it covers input devices
[00:40:05] <jmkasunich_> I'm guessing it should be called 85-input.rules
[00:40:56] <skunkworks> As a prerequisite you need a rule in the udevs (dynamic device management) rules folder (/etc/udev/rules.d, use sudo).
[00:41:14] <skunkworks> In my case it is a file called emc_joy.rules (the name can be chosen freely). There must be a line either for the use of various input devices or for a certain device. I used the variant for all devices (see manpages):
[00:41:47] <jmkasunich_> skunkworks: can you pastebin your file
[00:42:02] <jmkasunich_> I'll try it, if it works I'll instruct whatsisname to use it
[00:42:23] <skunkworks> I am just going by the wiki.. I have not actually done this.. :)
[00:42:36] <SWPadnos> the README suggests rules in the 40s
[00:42:40] <skunkworks> SUBSYSTEM="input", mode="0660", group="plugdev"
[00:42:43] <SWPadnos> for things that change device permissions
[00:42:51] <skunkworks> http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?A_New_Approach_For_Using_Joypads_With_EMC2
[00:43:07] <SWPadnos> or in the 50s for "user-supplied rules"
[00:43:22] <jmkasunich_> oh, sorry skunkworks - didn't realise that
[00:44:22] <skunkworks> you should have realised it.. I cannot sound that smart on a good day ;)
[00:45:29] <SWPadnos> ahhhh - much better :)
[00:45:55] <SWPadnos> I made a file called 50input-permissions.rules, with that single line in it
[01:11:19] <jmkasunich_> I should either fix hal_joystick (stop using the PID, and make it set hal_ready) or mark it deprecated
[01:12:04] <SWPadnos> or add an option to set the name, like pyvcp
[01:12:22] <jmkasunich_> yeah, that would allow multiple instances
[01:12:58] <jmkasunich_> hal_input handles things in a more elegant way (naming, multiple devices, etc), but the udev hoops are a pain for a newbie to jump thru
[01:13:12] <jmkasunich_> heck, they were a pain for me to jump thru
[01:13:49] <jmkasunich_> oh well, not gonna do anything tonight - leaving in a little while
[01:14:37] <skunkworks> glad you guys could figure it out.. Thanks
[01:15:03] <SWPadnos> I hadn't ever gotten around to actually making the one-line file
[01:15:20] <SWPadnos> even though I tried to debug some Tormach-related stuff (for that jog shuttle)
[01:15:28] <SWPadnos> I just ended up running as root
[01:15:38] <SWPadnos> thanks for that link skunkworks - it was a big help
[01:16:10] <skunkworks> np
[02:22:08] <skunkworks> some one is snookered up :)
[04:06:00] <cradek> hey folks, I have some very puzzling behavior in 2.2.6. I am getting infinite acceleration on jogs. It happens in sim/axis too.
[04:06:10] <cradek> can someone else look to make sure I'm not insane?
[04:06:18] <cradek> (I've been tuning for a long time tonight...)
[04:06:31] <SWPadnos> hmmm - plain old sim/axis?
[04:06:46] <SWPadnos> is 2.2.7~cvs close enough?
[04:08:13] <cradek> yes surely it is
[04:13:18] <SWPadnos> I see +/- 20 when I jog
[04:14:29] <SWPadnos> which is the ini setting
[04:15:06] <cradek> ok I must be confused. but for g1 I'm seeing accels of 4
[04:15:35] <SWPadnos> G1F10000 even?
[04:15:56] <cradek> F5, same as the jog
[04:16:14] <cradek> hm, accels are half in the tp, aren't they
[04:16:54] <SWPadnos> indeed - it's 10 for a G1F10000
[04:17:21] <SWPadnos> or G0
[04:17:42] <cradek> ok thanks, you confirmed that I'm insane
[04:17:56] <SWPadnos> excellent. good night! :)
[04:18:01] <cradek> goodnight
[04:18:18] <cradek> the lesson is don't look at a position plot and think you can read the acceleration from it
[04:19:12] <SWPadnos> heh - yeah, Xacc is probably better for that
[04:19:45] <cradek> well, I didn't have those in my hnc config
[04:20:22] <SWPadnos> what's the accel there?
[04:20:41] <cradek> 20 (10)
[04:20:47] <SWPadnos> ah, same as sim
[04:21:06] <cradek> goodnight :-)
[04:21:12] <SWPadnos> see you
[06:35:51] <CIA-36> EMC: 03seb 07TRUNK * 10emc2/src/hal/drivers/mesa-hostmot2/ (ChangeLog TODO hostmot2.h stepgen.c):
[06:35:51] <CIA-36> EMC: hostmot2 0.10:
[06:35:51] <CIA-36> EMC: Added stepgen.maxaccel.
[06:35:51] <CIA-36> EMC: This version of the stepgen position controller and velocity slewing
[06:35:51] <CIA-36> EMC: code is the least broken one yet.
[06:35:53] <CIA-36> EMC: Added stepgen.velocity_cmd (not writable, just for peeking at what
[06:35:57] <CIA-36> EMC: the velocity slewing code is doing).
[06:53:20] <CIA-36> EMC: 03seb 07TRUNK * 10emc2/docs/man/man9/hostmot2.9: describe stepgen.maxaccel and stepgen.velocity-cmd in the hostmot2 manpage
[06:53:30] <CIA-36> EMC: 03seb 07TRUNK * 10emc2/src/hal/drivers/mesa-hostmot2/TODO: describe stepgen.maxaccel and stepgen.velocity-cmd in the hostmot2 manpage
[11:57:59] <jepler> whee it's good to see accel in hm2
[12:56:27] <Lerman> Who is the Halscope guru?
[13:03:06] <cradek> if you ask your question you might find out
[13:12:42] <Lerman> I was wondering how hard it would be to add a user selectable configuration file to halscope.
[13:15:49] <Lerman> I have a config that I used to tune X, another to tune Y, a third to tune Z. Right now, I manually copy the one I want to .scope.cfg. It would be nice if halscope had buttons to load/save a config.
[13:16:21] <skunkworks_> how is the tuning going?
[13:16:26] <Lerman> Also to save a trace and to reload it.
[13:17:05] <Lerman> Tuning is going well. The Z axis has a little more error than I'd like +- .0005, but I could live with that.
[13:18:18] <Lerman> Halscope is a super tool. Jon Elson's tuning page in the wiki gives some pretty good directions. [But could use some additional stuff on using halscope for us neophytes.]
[13:19:13] <Lerman> That's one of the reasons I'd like to have loadable configs. We could ship them as part of a tuning package.
[13:34:27] <cradek> Lerman: I would love to have that feature too
[13:34:52] <cradek> ... and I thought someone wrote most of it already
[13:36:10] <cradek> yes, you can specify a config file on the halscope command line
[13:36:14] <Lerman> Well, it already can read and save a configuration file. So the only stuff that really needs to be written is the gui stuff. And that could be pretty much copied from the logging file selector.
[13:37:07] <cradek> I agree
[13:37:18] <cradek> I see this is new in trunk
[13:37:52] <cradek> I would like to see a menubar with at least a File and Help menu. I don't think adding more random buttons is the way to go.
[13:38:34] <Lerman> If I invoke it with a config file in the command line, will it save changes back in the file when it exits. I guess that doesn't really matter, though because it could be invoked from a script that copied .scope.cfg into the original file if that is what is wanted.
[13:38:51] <Lerman> I agree. File and Help would be clean.
[13:38:53] <cradek> 'halscope -i infile -o outfile'
[13:39:25] <Lerman> Also, I'd like to be able to dump the current screen display in such a format that it could be recalled. (Or is that what the logfile does?)
[13:40:01] <cradek> not sure what logging does, but I also want that feature :-)
[13:40:38] <cradek> all the time, we send around scope screenshots. you cannot zoom them. save/load of the data would be extremely useful
[13:40:41] <Lerman> All part of tuning. It would be nice to be able to compare the result of two tuning runs.
[13:41:07] <cradek> it would be cool if the save/load could use a format that could be loaded into a spreadsheet for analysis.
[13:41:19] <cradek> (might as well, since we have to make up a save format anyway)
[13:41:56] <Lerman> Did someone say XML?
[13:42:02] <Lerman> :-)
[13:42:17] <cradek> sorry, you must be mistaken, I heard comma-delimited
[13:42:44] <cradek> (I know nothing about spreadsheets and their file formats)
[13:42:46] <Lerman> You are probably correct. The hearing in my right ear isn't what it used to be.
[13:44:21] <Lerman> That format is easy. One line for each "signal" with the name and all of the values. It would probably be good to prefix the signal lines with the saved configuration.
[14:21:30] <jepler> creating a chart from 2000 rows of data takes many many seconds in oocalc (ubuntu 6.06)
[14:23:34] <jepler> but gnuplot has no trouble with that much data
[14:24:45] <jepler> http://emergent.unpy.net/files/sandbox/halscope-to-gnuplot.png http://emergent.unpy.net/files/sandbox/halscope-to-oocalc.png
[14:25:05] <jepler> (I chose to use space-separated instead of comma-separated because that's apparently what gnuplot likes)
[14:25:18] <jepler> I would check this in but I don't know who might have found a way to use the present logging format ..
[14:26:27] <cradek> hahaha oocalc's X axis is great
[14:26:54] <jepler> http://emergent.unpy.net/files/sandbox/scope_logging_ssv.patch
[14:28:26] <cradek> looks very useful to me
[14:34:31] <Lerman> I think we still need a format that can be read back by halscope. That would let us interactively adjust the scale, turn traces on and off, etc. We would still need what you've done to do analysis.
[14:36:42] <cradek> yes, reading it back in would be great.
[14:38:42] <Lerman> Quick. Go Hide. BigJohnT is here.
[14:38:48] <Lerman> :-)
[14:38:51] <BigJohnT> duck
[14:38:57] <Lerman> quack
[14:39:14] <BigJohnT> I'm not really here, I'm paying bills
[14:40:46] <jepler> seb_kuzminsky: I'm excited to see acceleration in the hm2 driver .. I look forward to giving it a try
[14:40:58] <seb_kuzminsky> cool :-)
[14:41:24] <jepler> ouch -- "In an email sent to the fedora-announce mailing list, it has been revealed that both Fedora and Red Hat servers have been compromised. As a result Fedora is changing their package signing key. Red Hat has released a security advisory and a script to detect potentially compromised openssh packages."
[14:41:38] <seb_kuzminsky> i'm not sure if the algorithm is right... in fact i bet it's got some bugs here and there
[14:44:38] <jepler> this is another thing it would be nice to see as a library item -- hostmot2, pluto_step, and ppmc all want to control step frequency with acceleration and velocity limits where steps are generated from a clock that is not exactly coupled to the PC clock. hm2 and pluto_step both have their own solutions for this, while ppmc uses an external pid loop which has to be tuned (it only has a velocity input, not position)
[14:45:18] <seb_kuzminsky> yes
[14:46:09] <jepler> oops, work calls -- bbl
[15:13:42] <seb_kuzminsky> cool, "bzr checkout svn+ssh://..." works :-D
[15:14:48] <seb_kuzminsky> now if it only would work with CVS i'd be all set ;-)
[15:24:00] <Lerman> Has anyone tried to build a hal component to do auto tuning of PIDs?
[15:27:03] <skunkworks_> there is a at_pid that uses the Ziegler and Nichols method. I don't know if anyone has had great luck with it.
[15:27:34] <skunkworks_> http://www.linuxcnc.org/docs/2.2/html/man/man9/at_pid.9.html
[15:27:57] <skunkworks_> I think it works better with velocity amps. (from what tom has mentioned)
[15:28:06] <Lerman> My guess is that it will give P values that are a lot smaller than I'm using. So it wouldn't be as "stiff".
[15:28:43] <Lerman> When I tuned, I went to the max P that would just oscillate. Then I increased D as far as I could. Then increased P some more.
[15:29:19] <Lerman> So the P will probably be twice that of a ZN tuning.
[15:32:21] <skunkworks_> when I played with it - the setting where squirley.. If that made sense
[15:32:51] <skunkworks_> (but usable for playing)
[15:32:54] <fenn_> fenn_ is now known as fenn
[15:33:21] <skunkworks_> I would think there would be a way to tune automagically but I don't have the smarts to do that.
[16:42:31] <CIA-36> EMC: 03cradek 07TRUNK * 10emc2/src/hal/utils/ (scope.c scope_horiz.c scope_usr.h): add a menubar and move the logging to it. make a help/about dialog. the things Ken and I were talking about this morning are on the menu but greyed out becuase they are not implemented yet.
[17:09:00] <SWPadnos> Lerman, I found a reference to some auto-tuning code that was written by an intern at NIST. I didn't see the code though.
[17:09:47] <SWPadnos> Lerman, jepler found an email in the old archives where the software was discussed, look at yesterday's log for the link
[17:15:15] <skunkworks_> I remember reading about that a few years ago.
[17:19:30] <SWPadnos> here's the start of the email thread: http://emc.sourceforge.net/NIST-archive/msg00455.html
[17:20:45] <SWPadnos> and here's the part of the "Handbook" with Fred's report: http://www.linuxcnc.org/handbook/part2/pidtune.html
[17:25:27] <skunkworks_> do you think that fred would still have the files?
[17:25:53] <SWPadnos> it's possible, though the files he refers to are Word docs that show how well the code worked
[17:41:43] <skunkworks_> http://www.yty.net/cnc/faq/faq/faq.pdf
[17:41:58] <skunkworks_> search for stolk
[17:42:52] <skunkworks_> around page 31
[17:46:14] <skunkworks_> http://bsing.ing.unibs.it/~visioli/didattica/
[17:46:24] <skunkworks_> pidtune.zip is on this page..
[17:46:31] <SWPadnos> cool
[17:46:53] <skunkworks_> I don't think it is right - but it looks like a neat little spreadsheet
[17:49:35] <skunkworks_> the pdf sorta gives how it was calculated.
[17:49:53] <SWPadnos> yep, sorta :)
[18:19:11] <cradek> I notice nobody has finished the halscope features I put on the menu yet...
[18:21:35] <alex_joni> heh, wishful thinking
[18:21:41] <cradek> darnit
[18:21:49] <alex_joni> wonder if anyone tried it out yet
[18:22:11] <cradek> I can probably pull off save and load config
[18:22:41] <cradek> the offline viewer thing is probably harder
[18:22:53] <alex_joni> run it through hal_streamer
[18:27:33] <SWPadnos> except that it requires scope_rt to load before it'll run, and halstreamer is also RT
[18:27:51] <SWPadnos> not that it's a problem for an initial effort ...
[20:13:33] <CIA-36> EMC: 03cradek 07TRUNK * 10emc2/src/hal/utils/ (scope.c scope_disp.c scope_usr.h): load and save of halscope configurations
[20:17:42] <skunkworks_> woo whoo!
[20:17:47] <skunkworks_> :)
[20:17:51] <cradek> thanks
[20:17:58] <cradek> but wait til you see how bad the gtk filechooser is...
[20:18:11] <skunkworks_> I will blame you ;)
[20:19:06] <cradek> good idea - it probably is my fault as I noticed there were many filechoosers to choose from (heh), but I didn't try them all
[20:20:57] <SWPadnos> let's write our own! :)
[20:21:13] <SWPadnos> like the signal chooser for hal{mneter,scope}
[20:21:16] <SWPadnos> -n
[20:21:21] <cradek> arg
[20:21:24] <SWPadnos> heh
[20:21:51] <SWPadnos> I feel your pain - remember, I'm one of the ones who tried to get the damned thing to allow incremental searching - what a PITA
[20:22:23] <cradek> oh yeah, wow, hey by the way that doesn't work right
[20:22:31] <SWPadnos> yes, I know :)
[20:22:49] <SWPadnos> which version of wrong do you prefer? the one in halscope or the one in halmeter?
[20:23:49] <cradek> hahaha
[20:23:50] <cradek> I dunno
[20:23:57] <SWPadnos> me either :)
[20:24:30] <cradek> I'd be pseudohappy with just pageup, pagedown, uparrow, downarrow
[20:27:34] <SWPadnos> don't those work?
[20:28:17] <SWPadnos> oh - they do, but the scroller doesn't let you see what's selected
[20:28:20] <SWPadnos> stupid scroller
[20:37:34] <SWPadnos> cradek, did you notice if that file chooser lets you add other buttons for quickly choosing a directory?
[20:37:44] <SWPadnos> like one for the nc_files dir
[20:58:54] <cradek> no I didn't notice, but that might be cool.
[20:59:09] <cradek> unfortunately it doesn't know how to show just *.halscope files, so I think it sucks
[20:59:18] <cradek> there is at least one other file chooser in gtk that might suck less
[20:59:40] <cradek> so I'd recommend trying that one before trying to add features to this one
[21:16:53] <micges> good night all
[21:52:36] <jmkasunich_> it sounds like gtk may not have been the best choice for halscope
[22:30:30] <jepler> jmkasunich_: oh I think all the gui toolkits suck
[22:30:40] <jmkasunich_> heh
[22:32:46] <mozmck> looks like Axis uses tk, right?
[22:34:36] <jepler> yes that's right
[22:35:20] <mozmck> is RTAI in the repository for ubuntu compiled with APIC support?
[22:35:52] <mozmck> jepler, have you looked at fltk? It is supposed to be fast - I don't know...
[22:36:59] <jepler> mozmck: I think that at least for 6.06, we chose not to use apic because it is not available on all machines. having a kernel and rtai that worked on the widest range of machines is pretty important since it lowers the barrier to new users
[22:37:15] <jepler> I don't have an 8.04/i386 machine to see what configuration we ended up with there
[22:37:27] <mozmck> on RTAI, I was reading here: http://www.captain.at/rtai-manual.php?rtai=5, which is why my question.
[22:37:32] <jepler> but being 2+ years later I think the equation has changed as far as stuff like the availability of apic is concerned
[22:38:44] <jepler> $ lsb_release -d; grep APIC /boot/config-2.6.15-magma
[22:38:44] <jepler> Description:Ubuntu 6.06.2 LTS
[22:38:44] <jepler> # CONFIG_X86_UP_APIC is not set
[22:39:40] <mozmck> I see. I'll check my 8.04 box later.
[22:40:18] <mozmck> mainly just curious. I turned it off thinking it was ACPI in the BIOS.
[22:40:52] <mozmck> The machine gives me a realtime error every so often.
[22:41:16] <mozmck> but I don't know that it has anything to do with apic
[23:06:46] <mozmck> CONFIG_X86_UP_APIC is not set
[23:07:01] <mozmck> that's from the 8.04 src file
[23:07:15] <mozmck> .config file i mean