#emc | Logs for 2008-03-05

[00:13:03] <lewin1> lewin1 is now known as lewing
[00:26:52] <Adam1> nick MASEngr\
[00:26:56] <Adam1> Adam1 is now known as MASEngr
[00:27:13] <MASEngr> *sigh* At least nobody saw that.
[00:28:02] <MASEngr> Hi everyone. I've got some trouble with my CNC machine. I get an axis following error when trying to mill.
[00:28:06] <BigJohnT> but I did
[00:28:32] <MASEngr> That's the joke, BJT. Everyone in the room saw the tpyo.
[00:28:50] <BigJohnT> yep
[00:29:12] <BigJohnT> stepper or servo
[00:29:29] <MASEngr> The pulses are showing up on the PC side of the connection (checked with a little pulse tester) and the display changes when the axis is cranked by hand.
[00:30:26] <tomp> are you describing encoder pulses?
[00:30:31] <BigJohnT> servo I assume
[00:30:44] <SWPadnos> or at least something with feedback
[00:31:00] <MASEngr> YEah, servo. AFK to check to make certain.
[00:31:17] <MASEngr> Yes, the linear encoder pulses are coming back fine.
[00:31:51] <BigJohnT> the question is not for me then
[00:31:52] <SWPadnos> how much backlash is there between the motor and the table/gantry?
[00:33:43] <MASEngr> Please wait. Getting backlash results...
[00:34:06] <SWPadnos> ok. just as a hint, the maximum measurable backlash is the minimum usable MIN_FERROR
[00:34:15] <SWPadnos> it may need to be higher
[00:34:42] <ds2> Mmm oscillation potentials
[00:34:51] <SWPadnos> that's another issue :)
[00:35:15] <SWPadnos> but spinning the motor and seeing no feedback = measured following error
[00:35:59] <ds2> does EMC do lash compensation in a stepper setup?
[00:36:26] <SWPadnos> it can, but you still need to deal with FERROR, unless you change the normal stepgen <-> motion controller connections
[00:36:35] <MASEngr> We can mill most of the time. There's an intermittent error where we lose an axis.
[00:36:48] <ds2> so the lazy people will just write lash compensated g code ;)
[00:37:12] <BigJohnT> is it during heavy cuts?
[00:37:17] <SWPadnos> MASEngr, what are your FERROR nad MIN_FERROR settings?
[00:37:36] <SWPadnos> ds2, hmmm. backlash compensated G-code. that's finny :)
[00:37:37] <MASEngr> It doesn't seem to make a difference.
[00:38:04] <ds2> SWPadnos: no G03/G02's and always feeding on a consistant direction... no worse then on a manual mill
[00:38:32] <MASEngr> FERROR and MIN_FERROR are both 0.05 on x and y, and .01 on z.
[00:38:34] <SWPadnos> don't forget that you always have to either climb or conventional cut - you can't mix
[00:38:41] <SWPadnos> MASEngr, mm or inches?
[00:38:49] <ds2> for some small mills, climb is not an option
[00:39:25] <MASEngr> Inches. Probably.
[00:39:36] <SWPadnos> that would be important to know ;)
[00:39:48] <ds2> whoa 0.050 for an error? Hmmmm
[00:39:58] <MASEngr> It says Linear = inches, and that axis is inches, so that's why I'm assuming inches.
[00:40:03] <SWPadnos> 0.05 inches is a large ferror, 0.05 mm may be in the same range as any backlash
[00:40:08] <SWPadnos> ok, it's inches then
[00:40:09] <MASEngr> Er, that axis is defined as linear.
[00:40:22] <ds2> wood working machine? ;)
[00:40:50] <MASEngr> Mostly aluminum.
[00:40:51] <SWPadnos> you're using software encoder counting, right?
[00:41:12] <MASEngr> If you tell me where I'd check that, I'd answer.
[00:41:18] <SWPadnos> actually, instead of dealing with this piecemeal, can you just put the ini file on http://pastebin.ca ?
[00:41:42] <MASEngr> I didn't set up the machine.
[00:42:03] <SWPadnos> the HAL files will tell you. if there's a loadrt encoder ... line anywhere, and some other lines that connect parport pins to encoder pins, then it's using software counting
[00:42:21] <SWPadnos> since you said the pulses were getting to the parallel port, it sounds like software counting
[00:42:40] <MASEngr> http://pastebin.com/d3d7ad9f5
[00:42:57] <MASEngr> That's the m5i20.ini file.
[00:43:07] <SWPadnos> and that's the one you use? :)
[00:43:17] <MASEngr> Yes.
[00:43:41] <SWPadnos> do you get faults on different axes, or is it always one axis?
[00:44:15] <MASEngr> It's always the x axis, except for one time when it was Y.
[00:45:27] <SWPadnos> it may just be a PID tuning issue, which I don't think I'll be able to help with (especially over IRC)
[00:46:43] <MASEngr> Oh, that's good news. No, wait. The other thing.
[00:47:16] <MASEngr> Oh, well, it doesn't matter, since I realized I have to leave immediately.
[00:47:25] <SWPadnos> see you later
[00:47:37] <MASEngr> Thanks anyway. I'd lost track of time, and the kids have to be picked up by ten to five.
[02:10:53] <eric_U> jjljlkjkljl
[02:12:08] <skunkworks> I tripped over the cable agin
[02:12:10] <skunkworks> again
[02:12:31] <eric_U> I knew it
[02:19:12] <gezar> how you guys doing tonight?
[02:55:07] <Guest713> Guest713 is now known as skunkworks_
[02:55:46] <skunkworks_> welcome back
[03:29:09] <gezar> I think I just broke my calculator, I fed it an equation to solve and its been 5 minutes
[03:51:00] <LawrenceG> gezar: that happens to my brain all the time :}
[04:06:45] <ds2> gezar: is that one that computes all the digits of PI?
[04:08:04] <ds2> and is the calculator called "NOMAD"?
[04:49:18] <ChanServ> [#emc] "This is the #emc channel - talk related to the Enhanced Machine Controller and general machining. Website: http://www.linuxcnc.org/, wiki at http://wiki.linuxcnc.org/"
[07:22:20] <ChanServ> [#emc] "This is the #emc channel - talk related to the Enhanced Machine Controller and general machining. Website: http://www.linuxcnc.org/, wiki at http://wiki.linuxcnc.org/"
[07:37:42] <[1]a-l-p-h-a> [1]a-l-p-h-a is now known as a-l-p-h-a
[09:53:55] <christel> [Global Notice] Hi all! We experienced some rather heavy ddos attacks earlier this morning causing quite a lot of disturbance to the network, the problems should be under wraps now (thanks to the OSL and NERO) and we apologise for any inconvenience caused. Hope you have a nice day and thank you for using freenode!
[12:38:50] <cradek_> cradek_ is now known as cradek
[14:00:25] <lerman> Any developers here?
[14:06:59] <fenn> dataja
[14:07:52] <cradek> lerman: try again
[14:07:56] <lerman> I've installed emc2 on a new machine and am trying to do a build. Can't seem to find libpth-dev.
[14:08:25] <cradek> Filename: pool/universe/p/pth/libpth-dev_2.0.1-2.1_i386.deb
[14:08:33] <cradek> you will have to enable universe
[14:08:55] <fenn> it's libpth20 on debian if that helps
[14:09:14] <lerman> Using apt-get what do I have to do?
[14:09:29] <cradek> first enable universe repository in sources.list
[14:09:31] <fenn> edit /etc/apt/sources.list and uncomment the universe repository
[14:09:35] <cradek> then apt-get install libpth-dev
[14:09:49] <fenn> apt-get update in there somewhere
[14:09:55] <cradek> oh right
[14:10:08] <cradek> deb http://us.archive.ubuntu.com/ubuntu/ dapper universe multiverse
[14:10:08] <cradek> deb-src http://us.archive.ubuntu.com/ubuntu/ dapper universe multiverse
[14:10:24] <alex_joni> hi guys
[14:10:35] <cradek> Mar 5 08:09:25 cvs sshd[27660]: Accepted publickey for lerman
[14:10:36] <cradek> whee
[14:10:50] <lerman> Thanks, cradek.
[14:10:54] <cradek> welcome
[14:13:00] <lerman> got libpth-dev. Thanks again.
[14:14:11] <lerman> ./configure --enable-run-in-place --enable-simulator
[14:14:27] <lerman> configure: error: Required header Python.h missing. Install it, or specify --disable-python to skip the parts of emc2 that depend on Python
[14:14:38] <cradek> apt-get build-dep emc2
[14:16:14] <lerman> woof. tetex-base 22.4MB -- there's a lot of stuff required.
[14:16:32] <fenn> and about a dozen translation packages
[14:16:37] <fenn> for texlive
[14:16:37] <cradek> yeah since the docs are in there too, there's a bunch of crap needed
[14:16:50] <fenn> i really think docs should be a separate package
[14:17:06] <lerman> That's OK. But I only have a 40Gig disk on this machine. :-)
[14:17:23] <cradek> lerman: 4G is still plenty :-)
[14:17:46] <lerman> It's an old machine, though. So I should replace the old disks before they crash, anyway.
[14:18:20] <jepler> if you're not actually building a package, you can intelligently remove some things from the list of packages proposed by apt-get build-dep
[14:18:27] <lerman> The machine is a dual pentium 1.4GHz 2Gig ram rack mount that I got for $150. (Actually, I have six of them)
[14:18:42] <jepler> but on the other hand, the list of packages proposed should be an absolutely complete list of the things required to rebuild the package
[14:18:58] <lerman> "intelligently" was the key word. And where would I get that from? :-)
[14:19:10] <jepler> making the documentation a separate binary package wouldn't help this, because it would still be built from the same source pakage
[14:19:14] <jepler> +c
[14:19:37] <jepler> lerman: for most of us, bandwidth is cheaper than smarts
[14:19:56] <cradek> I don't even recall where to buy smarts
[14:20:07] <lerman> I shouldn't complain about tetex. I use emacs as my preferred editor. I haven't downloaded that, yet.
[14:21:08] <jepler> I don't love the lyx "toolchain" we've got for our documentation, but on the other hand we've got a hell of a lot of words in .lyx format, so what are you going to do ..
[14:21:18] <fenn> apt-get wants to install 128 new packages even though i can compile fine already
[14:21:26] <fenn> that's a complexity issue, nothing to do with bandwidth
[14:21:50] <cradek> fenn: whether you can compile is not important. whether you can build the package that matches the binary we distribute is
[14:22:12] <fenn> sure, and i think the docs should go in a different binary package
[14:22:13] <cradek> that's what build-dep is for. it's convenient that we can use it for other things.
[14:22:43] <cradek> brb
[14:22:46] <lerman> Does the toolchain we distribute let me build the ubuntu iso file?
[14:23:04] <fenn> lerman: he means the .deb file
[14:23:27] <lerman> I don't know what a .deb file is.
[14:23:41] <fenn> ah err.. maybe it's time to learn then
[14:24:06] <lerman> Only if I need to know. Otherwise you might have to shoot me.
[14:24:24] <fenn> deb files are what 'packages' come in, the things you install with apt-get
[14:24:47] <lerman> I figured. But how do I build an 'iso'?
[14:24:50] <fenn> there are other packaging systems, like rpm or puppy or slackware
[14:25:09] <fenn> i dont know how to build a live-cd, presumably the ubuntu wiki can teach you
[14:25:17] <fenn> a .iso is just an image of a cd rom
[14:25:27] <fenn> short for iso9660
[14:25:46] <lerman> Yup. I know that. Who is it that builds our iso files, though?
[14:26:03] <lerman> (And do we have a backup?) ;-)
[14:26:13] <fenn> i'll let the guilty incriminate themselves :)
[14:32:38] <fenn> lerman: making the livecd is only part of it, you also have to make sure the realtime kernel works for most computers
[14:34:07] <lerman> The actual question is simply this. Given access to all of the source, makefiles, etc, how does one reproduce the emc2.2.xxx...iso file that I just installed?
[14:34:53] <lerman> There should be a written procedure for doing this. Since the original was tested, I shouldn't need to do any.
[14:35:57] <lerman> It would be nice to be able to automatically generate one for each minor release of emc2. Presumably, the kernel would be unchanged.
[14:36:26] <skunkworks> Normally alex does it.. :)
[14:36:50] <skunkworks> but yes - there isn't a 2.2.3 iso yet.
[14:36:52] <lerman> I'll ask him sometime. It's really just curiousity.
[14:38:33] <lerman> OK... my make completed without error. Now that I have a development environment, I can pretend to be a developer again.
[14:38:44] <skunkworks> about time... ;)
[14:39:43] <lerman> Indeed. The first step is to test this in simulation (I don't have a machine connected). Then I can go to sourceforge and find the bugs with my name on them. It is about time I got around to fixing them.
[14:40:59] <fenn> welcome back
[14:43:05] <lerman> tnx.
[14:43:24] <cradek> I disagree 100% that a new iso should be generated for every little release
[14:43:36] <cradek> that leads to the nightmare of users thinking they should update by reinstalling their OS.
[14:44:02] <cradek> fenn is right that the procedure to make a custom ubuntu cd with changed packages is well covered in the ubuntu docs (wiki)
[14:47:41] <cradek> aside from that, welcome back lerman :-)
[14:48:37] <lerman> Thanks. I was just curious about how one does that. I didn't mean to imply that we should generate a new iso for each release -- only that we should be able to.
[14:49:07] <cradek> it's really quite easy to change the packages on the cd. the ubuntu folks get all the credit for that of course
[14:49:13] <lerman> I went to the web site, found out that the current release was 2.2.3 (I think), went to install it, and found that I couldn't.
[14:49:24] <cradek> couldn't what?
[14:49:42] <lerman> I couldn't just install a 2.2.3 from CD.
[14:50:11] <lerman> Not that it really matters. I have to be able to build the latest from source, anyway.
[14:50:20] <cradek> when I was making the CDs I purposely did not put the emc version number in the filename. Alex seems to disagree with that.
[14:50:47] <cradek> lerman: a normal update would install the latest version. there's even an icon on the screen that says updates are available.
[14:50:58] <lerman> (If you want it done right, you have to do it yourself.) :-)
[14:51:10] <lerman> Yup.
[14:51:13] <cradek> heh, well there's probably more than one "right"
[14:51:23] <lerman> Of course.
[14:51:34] <lerman> I just found a bug with my name on it.
[14:51:36] <cradek> but saying which version is on the cd led you to believe that you were getting something wrong
[14:51:45] <lerman> [ 1807761 ] Local named parameters not scoping correctly in subroutines
[14:51:55] <cradek> I think there are several waiting for you :-)
[14:52:21] <lerman> The problem with this one is that named o-words don't work right. But I don't think that I ever implemented that. Do we know who did?
[14:52:37] <cradek> hmm I figured it was you...?
[14:52:44] <cradek> just check cvs history
[14:53:20] <lerman> I don't think so. I seem to remember someone doing that using some STL functionality. I guess I'll have to read the source.
[14:53:57] <lerman> But first I have to install emacs. :-)
[14:53:58] <cradek> http://cvs.linuxcnc.org/cvs/emc2/src/emc/rs274ngc/interp_o_word.cc?rev=1.3
[14:54:13] <cradek> Sat Dec 30 23:12:04 2006 UTC (14 months ago) by lerman
[14:54:31] <cradek> maybe you were in a year-end daze
[14:55:37] <lerman> That was named parameters. It should not include named o-words. I'm sure they never worked for me. It could be that I installed part of the infra structure without disabling the functionality.
[14:55:59] <lerman> I'm pretty sure the docs say not yet implemented.
[14:56:03] <cradek> oh sorry, I guess I didn't read all the words
[14:57:50] <lerman> It drives me nuts when people submit anonymous bug reports. It's hard to ask follow up questions.
[14:58:03] <cradek> I wholeheartedly agree
[14:58:22] <cradek> I think sf makes it easy to do that accidentally even
[15:03:23] <lerman> Hey. There are only three bug reports with my name on them. The first one I'll fix is cradek's.
[15:03:28] <lerman> [ 1709015 ] comments in O-call lines disrupt parameters
[15:03:32] <cradek> cool
[15:03:43] <cradek> since then, I noticed that comments disrupt other things too
[15:03:53] <cradek> g0 x (boo!) 3
[15:04:24] <lerman> That's probably a language definition thing.
[15:04:30] <cradek> so, I'm not so sure it's a big deal
[15:04:49] <cradek> g0 (this is allowed) x3
[15:04:56] <cradek> g0 x (this is an error) 3
[15:05:12] <lerman> Does it generate an error?
[15:05:15] <cradek> yes
[15:05:29] <cradek> I think it's "mumble mumble reading real number"
[15:06:10] <lerman> Yup. When the parser sees the X, it calls the read real number routine. That routine probably doesn't allow comments.
[15:06:33] <lerman> g0 x1.2(comment)345
[15:06:42] <lerman> should that be allowed?
[15:07:18] <cradek> I'm trying to find that in the spec
[15:08:38] <cradek> I do not see anything about exactly where comments are allowed
[15:08:48] <cradek> it doesn't say in the "Comments" section, or the "Format of a Line" section
[15:09:25] <lerman> How about in the definition of a "word"?
[15:09:58] <cradek> A word is a letter other than N followed by a real value.
[15:10:16] <cradek> I was wrong, I think "Format of a Line" does say
[15:10:33] <lerman> To me, that means that comments are not allowed in the middle of a word.
[15:10:56] <cradek> I agree
[15:11:06] <cradek> are you looking at http://www.isd.mel.nist.gov/personnel/kramer/pubs/RS274NGC_3.web/RS274NGC_33a.html#1001301
[15:11:10] <cradek> #3 in that list
[15:11:24] <cradek> (brb)
[15:13:50] <lerman> Appendix E is pretty clear that comments are not allowed in the middle of a word.
[15:15:28] <fenn> openfarmtech.org looks really neat
[15:15:38] <fenn> surprised i hadn't heard of it before
[15:29:38] <jepler> cradek: I think there's a way to disable anonymous bug reports; my recollection is that the python project did that back when their bug tracker was on sf
[15:39:30] <Guest697> Guest697 is now known as skunkworks_
[15:43:50] <fenn> Dallur: did you know that there is (soon to be) a fab lab in reykjavik?
[15:45:06] <lerman> So... I fixed a bug (perhaps). Did I break anything else? Do we have a test suite?
[15:45:35] <Dallur> Fenn: What ?
[15:45:51] <lerman> Since the bug is the comment one reported by cradek, the fix shouldn't break anything not related to call's.
[15:46:08] <lerman> So this one is easy to test.
[15:46:34] <Dallur> Fenn: hmm thanks for bringing that to my attention :)
[15:46:57] <fenn> Dallur: i havent found much else in a quick googling, this is what i'm referring to: http://smari.yaxic.org/blag/category/general/fab-labs/
[15:46:57] <micges> hello all
[15:49:42] <Dallur> Fenn: I think it might have been 2 weeks ago though :P
[15:50:11] <fenn> oh hey you are on that page :)
[15:56:07] <Dallur> Fenn: Lol Smari works for a company in the same street as my company, I know most of his coworkers to, small world
[15:57:46] <fenn_> fenn_ is now known as fenn
[16:06:03] <cradek> lerman: there is a test infrastructure. There are several interpreter tests you can run. I don't think any are related to O words; I can help you add the test cases you're interested in if you like.
[16:07:04] <skunkworks> wait - I think you could be a little more vauge.. but not much. http://www.cnczone.com/forums/showthread.php?t=53920
[16:10:51] <cradek> I was wrong: there is one O word test already.
[16:11:35] <cradek> (although I don't understand what exactly it's testing...)
[16:20:51] <BigJohnT> I know what went wrong, my advice is to call Gamma-X
[16:25:08] <skunkworks> To whoever moved the search dialog box to the top of the wiki page - thank you.
[16:25:49] <archivist> I wonder if Gamma=x ever cut metal yet
[16:32:15] <jymm> skunkworks: Next week it'll be on the side, then you'll have to learn to type/read veritcally =)
[16:32:35] <jymm> s
[16:32:35] <jymm> e
[16:32:35] <jymm> a
[16:32:36] <jymm> r
[16:32:36] <jymm> c
[16:32:37] <jymm> h
[16:33:50] <BigJohnT> archivist: I wonder too what happened to Gamma-X
[16:34:34] <BigJohnT> did you buy the lathe for 50 pounds?
[16:35:00] <archivist> no not up yet
[16:35:29] <archivist> 3 days to go
[16:35:54] <archivist> * archivist cant afford much, stupid job
[16:47:36] <Adam1> \nick MASEngr
[16:47:43] <Adam1> Adam1 is now known as MASEngr
[16:48:28] <MASEngr> Good morning, everyone.
[16:49:03] <MASEngr> I'm having some trouble with the CNC machine here. I'm using EMC2 v 2.2.3 and we're getting a lot of axis following errors.
[16:50:07] <MASEngr> When we disconnect the wires with the encoder, the pulses present.
[16:51:01] <MASEngr> The EMC2 window shows that the changes are being read on the computer - in other words, a physical change in position corresponds to a change in the display.
[16:51:49] <MASEngr> ANy suggestions?
[16:52:41] <skunkworks> is this the analam control conversion with scales?
[16:53:33] <MASEngr> It's an anilam machine which has been upgraded from crusader control to emc2, if that's what you mean.
[16:54:54] <MASEngr> If by "the" you mean you remember Adam, he moved to Bermuda six months ago and I wasn't part of the original setup.
[16:55:34] <skunkworks> yes. what the heck? bermuda?
[16:56:15] <MASEngr> Yeah, with the tax rate there, he's making the equivalent of about $200k / year here, once you consider take-home pay.
[16:56:17] <skunkworks> I think SWPadnos last night asked about the backlash.. If it has been increasing. IIRC adam thought there was a few thousands of backlash already in the setup.
[16:56:36] <skunkworks> yikes
[16:57:20] <MASEngr> That's right, SWP asked about the backlash and the PID tuning. I had to leave and pick up the kids, so I didn't get a chance to get anything checked out last night.
[16:58:22] <MASEngr> I don't know if the backlash has been increasing. The machine is already quite old and has been used heavily.
[17:00:42] <skunkworks> Just so you know - As the backlash increases - it is going to get to the point where you will not be able to tune it anymore. (using external linear scales)
[17:00:48] <MASEngr> Given the FERROR rates of 0.05 on X and Y, it was assumed that it's using a software encoder. As stated before, I am not sure how the machine is set up.
[17:02:08] <MASEngr> While Adam did a pretty good job, I think his focus was more on Bermuda than on documentation. Add to that the fact that I haven't used the software before, and you've got a recipe for disaster.
[17:02:48] <MASEngr> Well, okay, more inconvienience than disaster, but the general idea is still sound.
[17:03:50] <MASEngr> So what would that mean, SW? WOuld I have to replace the motors when that happens?
[17:04:04] <BigJohnT> http://www.cnczone.com/forums/showthread.php?t=52808
[17:09:09] <ChanServ> [#emc] "This is the #emc channel - talk related to the Enhanced Machine Controller and general machining. Website: http://www.linuxcnc.org/, wiki at http://wiki.linuxcnc.org/"
[17:09:09] <SWPadnos> FF1=2.5 seems strange, but that's probably due to the OUTPUT_SCALE
[17:09:09] <MASEngr> I love scopes too. They give the answers to all of life's questions.
[17:09:28] <SWPadnos> and -1.4 for D seems high as well (looking at AXIS_0 still)
[17:09:52] <cradek> uh negative D?
[17:10:21] <SWPadnos> on both X and Y, though Y is smaller magnitude
[17:10:31] <cradek> that's surprising to me
[17:10:36] <SWPadnos> almost nothing actually, -0.0001
[17:10:47] <cradek> I suspect plain old bad tuning
[17:11:09] <SWPadnos> seems that way, but I'm also pretty sure this is using linear scales for feedback
[17:11:19] <SWPadnos> so there could be other problems with "good" tuning
[17:11:23] <skunkworks> That is what I remember.
[17:11:24] <cradek> scales only?
[17:11:26] <skunkworks> yes
[17:11:35] <cradek> is there tach feedback to the amps?
[17:11:39] <SWPadnos> maybe not on Z though
[17:11:39] <skunkworks> yes
[17:11:54] <cradek> so the amps are velocity mode - that's good
[17:12:00] <cradek> I think you'd be stuffed otherwise
[17:12:09] <skunkworks> IIRC he use the cruser amps and scales.
[17:12:24] <skunkworks> net spit
[17:12:35] <MASEngr> Yes, they are using linear scales for position on X and Y.
[17:12:49] <cradek> this may be a hard machine to tune
[17:13:13] <skunkworks> iirc we told adam that ;)
[17:13:32] <skunkworks> He had decent results though - following error wise
[17:13:38] <cradek> I kind of doubt it will be able to actually go 3ips.
[17:14:25] <SWPadnos> considering that FERROR and MIN_FERROR are set to 0.050, I'm nos so sure the results were that good
[17:14:40] <cradek> yeah that's true
[17:14:50] <MASEngr> Last night, it was thought to be a wood-working machine.
[17:14:53] <cradek> that's a lot of error
[17:16:37] <MASEngr> That's inches.
[17:17:18] <cradek> that allows the machine to come to a stop .050" away from the commanded position and still not cause an error
[17:18:52] <MASEngr> Wait. What? That's more than a mm.
[17:19:12] <cradek> yes
[17:19:37] <cradek> it could be covering up pretty bad tuning
[17:19:55] <MASEngr> Yes, it could.
[17:20:02] <cradek> are you getting following errors at low or high speed? or whenever you move?
[17:20:20] <MASEngr> It's been intermittent. It's now "whenever we move".
[17:20:40] <cradek> whenever you move any axis?
[17:21:07] <MASEngr> It's been mostly on the X axis. There's only been one Y failure, and zero Z failures.
[17:21:28] <archivist> measure backlash
[17:22:49] <MASEngr> I'd just finished eliminating electrical causes, lilke loose wires, damaged connections, etc. I was about to move into checking for ground loops.
[17:23:22] <cradek> often if there is erratic counting, you can see it in halscope
[17:23:51] <cradek> for instance the count may sometimes change by 3-4 at a time instead of 1
[17:24:02] <MASEngr> How do I measure the backlash, and what channels would give the best info on halscope?
[17:24:22] <skunkworks> 16:58:55 <Adam1> I have tach feedback that goes to my drives
[17:24:31] <SWPadnos> I'd just start the tuning process over
[17:24:32] <skunkworks> 17:00:10 <cradek> how much backlash are we talking about?
[17:24:41] <skunkworks> 17:00:26 <Adam1> around 3 thou
[17:24:48] <cradek> ouch
[17:25:23] <SWPadnos> it's a good idea to set up the OUTPUT_SCALE so that an output of 1 from the PID makes the motor move 1 inch/second
[17:25:28] <cradek> MASEngr: simplest test is just push the axis side-to-side and measure it
[17:26:15] <skunkworks> http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2007-09-24.txt
[17:26:58] <MASEngr> That "cradek test" shows a difference on X of 0.0004
[17:27:39] <cradek> that sounds pretty good actually
[17:27:54] <cradek> that may be missing a few backlash sources, but probably gets most of them
[17:28:42] <tomp> you might set an indicator to 0, then turn the motor and record the scale feedback untill the indicator changes. the units that cause no motion = backlash ( done with hal reading encoder, but not with amplifier energized )
[17:28:48] <skunkworks> <Adam1_> SWPadnos this is my X-axis Error at 150ipm http://imagebin.org/10671 (scale is 1m), so about 0.001in at constant velocity and 0.002in at acceleration.
[17:29:12] <cradek> darn, the image is gone
[17:29:20] <skunkworks> images are gone :(
[17:29:35] <MASEngr> Hang on, another push with a little better grip gave me a 0.024 difference.
[17:29:46] <MASEngr> Er, 0.0024, sorry.
[17:29:55] <skunkworks> http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2007-09-27.txt
[17:30:21] <skunkworks> heh <fenn> Adam1_: 0.002" following error at 150 ipm is pretty good
[17:30:28] <skunkworks> <Adam1_> yeah, we arent gonna get too much bettter
[17:30:45] <cradek> MASEngr: that's the remaining error after you let go?
[17:30:53] <tomp> springiness is not backlash, does the position repeat? ( does the indicator read same before and after the push? )
[17:31:17] <cradek> yes push left, let go, set zero. push right, let go, read dial
[17:31:44] <MASEngr> I pushed it one way and held it, then pushed it the other way and held it, wrote down both values and subtracted.
[17:32:20] <MASEngr> Heh, I messed up the simple test.
[17:32:25] <cradek> ok, you're measuring the spring then.
[17:33:21] <cradek> a full-size guy can move just about anything .002 :-)
[17:33:25] <tomp> ( tho springiness is not good, you have to push as hard as the loaded system will push , hard for a hand and brain to do )
[17:34:08] <tomp> 'lean on me' (old song)
[17:34:36] <MASEngr> Redid the test: Pushed one way, let go, set X to zero. Pushed the other way, let go, and the value on X is: 0.0028
[17:35:16] <cradek> skunkwork> 17:00:26 <Adam1> around 3 thou
[17:35:29] <tomp> .0001" per division ? repeat test a few times for average & to insure not sticky
[17:35:47] <cradek> ok, sounds like about what Adam measured
[17:36:37] <fenn> if it's acme screws you might be flexing the threads in the anti-backlash nut
[17:37:08] <cradek> surely it's not acme!?
[17:37:28] <tomp> (shades of wiley e coyote)
[17:37:38] <MASEngr> 28, 39, 43, and I'm done pushing.
[17:38:34] <cradek> .0043?
[17:38:50] <MASEngr> Yeah, 0.0043.
[17:38:53] <archivist> i also test by dti and screw rotation
[17:39:14] <tomp> if it grows consistantly, i question the indicator more. if you dont, check everything is mounted tight, should not grow.
[17:39:34] <MASEngr> Yeah, it should get smaller as I get more tired.
[17:39:42] <cradek> heh
[17:39:42] <archivist> yup indicator has to be tight and good
[17:39:56] <alex_joni> MASEngr: try doing it for 5-6 hours
[17:40:11] <alex_joni> eventually it won't have any backlash, and you fixed the machine :P
[17:40:17] <tomp> you can wrap a cord around the pulley and use a fish scale for consistant torque
[17:40:30] <cradek> if X is working less and less well, it sure is possible there is a mechanical problem
[17:40:31] <skunkworks> sounds like a good job for a high school student
[17:40:39] <MASEngr> I don't have a pullet or a fish scale. Otherwise a great idea.
[17:40:53] <cradek> a loose pulley, motor mount, bad belt, who knows
[17:40:55] <MASEngr> 31 this time.
[17:41:28] <MASEngr> Funny you'd mention that, SW. This used to be a high school machine that they got rid of.
[17:42:12] <MASEngr> lol, Alex. I'd have trouble explaining why there are still following errors.
[17:42:27] <alex_joni> well, that's a proplem :P
[17:45:45] <MASEngr> All right. I'm going to go RTFM and see if I can figure it out from there. It sounds like there's a large error based on the physical setup of the machine, and the PID tuning maight be off, but covered slightly by the 1.3mm FERROR values on the X axis. If I can't solve the problem with retuning the PID loop, then I'll have to do something wacky like replace the scales.
[17:46:11] <MASEngr> It sounds like my week got a whole lot more interesting than I'd been planning.
[17:46:40] <SWPadnos> heh
[17:46:54] <cradek> if you need the machine to work, and make wood parts, just unhook the scale and put an encoder on the motor or ballscrew. That is virtually a sure thing.
[17:47:18] <cradek> .0043 in wood surely makes no difference
[17:47:29] <SWPadnos> I think wood was brought up by someone else, because of the large FERROR settings
[17:47:42] <cradek> oh
[17:47:43] <SWPadnos> ie "that's gotta be for wood with that kind of FERROR"
[17:47:47] <MASEngr> It's for Aluminium, for parts that are carrying people over the water.
[17:48:04] <SWPadnos> you probably want better than 0.050 then ;)
[17:48:37] <MASEngr> Yeah, that's what I was thinking.
[17:48:50] <MASEngr> It kind of explains the problem the machinist had with the machine.
[17:49:13] <fenn> what kind of parts, if you dont mind indulging my curiousity
[17:50:18] <MASEngr> I can't say exactly, but our tolerances are a tighter than 1mm.
[17:50:29] <MASEngr> Off I go. Thanks for the help.
[17:52:26] <fenn> * fenn pictures cranes lowering captured spies into vats of acid
[17:52:42] <SkinnYPuppY> Ha
[17:52:48] <tomp> aluminum walk-on-water shoes
[18:13:43] <skunkworks> sharks with laaaseeer beams
[18:22:57] <skunkworks> after I answered the question.. The realtime kernel shuts of the autoshutdown - correct? (you have to push the power button to shut the machine off)
[18:24:33] <alex_joni> skunkworks: the RT kernel has ACPI disabled
[18:24:43] <alex_joni> and ACPI is needed to halt the machine automagically
[18:24:51] <alex_joni> e.g. not having to push the button
[18:25:36] <skunkworks> Thanks
[18:25:45] <skunkworks> this still makes me giggle
[18:25:46] <skunkworks> http://www.electronicsam.com/images/eme.png
[18:26:17] <alex_joni> hehe, cool
[18:26:33] <alex_joni> did you try a wc -l on that file?
[18:27:14] <skunkworks> ? no. It was created from this jpeg http://www.electronicsam.com/images/me.JPG
[18:27:29] <alex_joni> skunkworks: I was wondering how many lines the g-code has
[18:27:57] <alex_joni> did you try to mill it?
[18:28:04] <skunkworks> ah - don't know - it was generated from emc2 image-to-gcode.
[18:28:20] <alex_joni> yeah, I figured that :P
[18:28:28] <skunkworks> Yes - in some plastic. didn't work the greatest as it was stringy. looked ok though
[18:28:42] <alex_joni> don't suppose you have a pic?
[18:28:58] <skunkworks> no - I would have to find it. I don't know if I kept it.
[18:29:07] <alex_joni> don't sweat it
[18:29:14] <alex_joni> gotta run anyways..
[18:29:32] <skunkworks> I was planning on making a video of it when I was selling the engraver. Never go to that though
[18:29:45] <alex_joni> too bad you sold it ;)
[18:30:17] <alex_joni> bbl.. gone for some exercise
[18:30:22] <skunkworks> have fun
[18:30:26] <alex_joni> thx
[18:50:10] <Adam1> Adam1 is now known as MASEngr
[18:50:23] <MASEngr> Hi everyone. Me again.
[18:51:36] <MASEngr> I'm having a slight issue with HALScope. I'm looking at the output pins for X and Y, but they are showing different scales. Is that normal?
[18:51:44] <MASEngr> http://imagebin.org/14597
[18:52:05] <MASEngr> Y is almost non-existent, whereas X is showing a strong pulse.
[18:58:01] <skunkworks> you can see by the tool tip - one is set to x is set 1/div and y is set to 100m/div.
[18:58:16] <skunkworks> You can click on each trace and change the scale to what you want.
[18:58:45] <skunkworks> 'gain'
[18:58:56] <MASEngr> Oh, yeah, I know that. I'm curious as to why the two signals are at such different levels.
[18:59:47] <skunkworks> oh heh
[19:00:09] <MASEngr> The pulse height on X is about .5, whereas there's almost no pulse height on Y. Both axes are moving.
[19:01:36] <MASEngr> I'm doing a dummy run with circles.
[19:02:01] <MASEngr> Of course, the stupid thing is working this morning.
[19:03:47] <MASEngr> And if I could do a quick sanity check, each axis could have drastically different PID parameters, right?
[19:08:01] <cradek> MASEngr: I don't understand what I'm seeing. What is Xoutput?
[19:10:25] <MASEngr> According to coreservo, it's linksp Xoutput <= pid.0.output
[19:11:12] <cradek> ok
[19:11:19] <cradek> what kind of motion is this?
[19:12:19] <MASEngr> I'm not quite sure how to answer that. Can you restate the question?
[19:12:40] <cradek> in what way was the machine moving while you made this plot?
[19:13:05] <MASEngr> Oh, it's making a circle.
[19:13:55] <MASEngr> And of course, it's _working_ this morning.
[19:14:36] <cradek> ok I only see .1 second on the plot so I did not see the circle.
[19:15:13] <cradek> I like to tune pid by moving an axis back and forth. It should be a long enough move to get up to accel, get up to cruise speed, then decel to a stop
[19:16:00] <cradek> the X output on your plot looks fishy to me. But the ferror does stay fairly low I guess.
[19:16:20] <MASEngr> I'm checking out halscope first to see what the baseline is before I start messing with things.
[19:16:55] <cradek> yes understand what it's doing, THEN make changes. so many people get the order confused :-)
[19:17:24] <MASEngr> But yeah, I agree that the X output seems kind of messed up. I was under the impression that a nicely tuned PID loop would give a smoother kind of graph instead of a bunch of pulses.
[19:17:33] <cradek> yes I agree
[19:18:15] <cradek> not sure what the scale is on your pid/amps, but if it's something like inches/sec, 0.6 is quite a large command
[19:18:37] <MASEngr> And if I don't measure first, I won't know if I've fixed the problem, masked it, or made it worse.
[19:18:43] <cradek> yes
[19:18:48] <MASEngr> (or done nothing)
[19:19:14] <cradek> hey one thing you might check is your tach feedback. Use a dual trace scope in difference mode on the amp's tach inputs and see if it looks smooth.
[19:19:57] <cradek> if they're just "extra" brushes in your motors maybe they are dirty or worn. that could cause the degradation and intermittent problems you've been seeing
[19:20:25] <tomp> any use? real scope and amp and battery box http://support.motioneng.com/Downloads-Notes/Tuning/pid_overshoot.htm
[19:20:33] <MASEngr> Heh, my real life scope doesn't do that.
[19:20:54] <cradek> no invert button on channel 2?
[19:21:07] <cradek> usually it's invert channel 2 and then add
[19:23:59] <MASEngr> Yeah, brain misfire.
[19:24:10] <cradek> just another thought
[19:24:15] <cradek> so many things to check :-)
[19:25:11] <MASEngr> It's an old 2213A, so I"m just used to saying "it doesn't do that".
[19:25:42] <archivist> even my cheap Gould does that
[19:26:26] <cradek> MASEngr: mine's a 466, it does everything I need (so far)
[19:26:38] <skunkworks> * skunkworks throws and isolation transformer at archivist
[19:27:32] <skunkworks> our old b&k I think also does add/invert. I can't think of the last time I used it though
[19:27:40] <archivist> skunkworks, the mains switch has failed and its always on, I have to unplug
[19:28:02] <skunkworks> hdh
[19:28:04] <cradek> you only have to clip the ground to the wrong thing once, to teach you to use add/invert
[19:28:04] <skunkworks> heh
[19:28:23] <MASEngr> So, do you ever have one of those days when you open the case and your day gets worse right away?
[19:28:35] <cradek> MASEngr: that's why we make the big bucks
[19:28:49] <skunkworks> I did that in collage.. Put rectifier bits in my hair and shirt pocket.
[19:29:35] <MASEngr> Once I blew a hole in a drop ceiling. That was a memorable lab.
[19:30:09] <cradek> "oh, ohms law. it'll be a straight line." <- summary of EE labs
[19:30:29] <MASEngr> It was a capacitor. Went straight up.
[19:30:37] <MASEngr> There was paper everywhere.
[19:30:40] <MASEngr> Everywhere.
[19:30:43] <cradek> whee
[19:30:52] <cradek> old radios are fun that way too
[19:31:07] <MASEngr> Yep.
[19:31:20] <cradek> oh look, a 450V, 70 year old capacitor
[19:31:27] <cradek> wonder if I should replace it
[19:31:28] <skunkworks> http://jmkasunich.dyndns.org/pics/CapBang1.jpg
[19:31:28] <archivist> hehe My first job was radio tv repair
[19:31:48] <cradek> nice pic
[19:31:48] <MASEngr> Ah, before they started putting resistors across the terminals.
[19:32:13] <MASEngr> Wow. Any survivors.
[19:32:14] <cradek> field coil speakers!
[19:32:17] <archivist> good one, they dont often come out the side
[19:32:42] <MASEngr> Anyway, given my last statement about opening the case, I have a fair bit of work to get to before I can get to work.
[19:32:55] <archivist> had a capacitor can fly off a pcb past my ear
[19:33:12] <tomp> this is nice from kollmorgen. using a sqr wave http://www.pid-tuning.com/?gclid=CM6mtMnZ9pECFQMKPAodjjqgxA
[19:33:20] <MASEngr> I was in a lab with a guy who got the die from an IC blow out of the case and stick into his forehead.
[19:33:45] <cradek> wow
[19:33:49] <MASEngr> It was, according to him, [expletive] hot.
[19:34:03] <cradek> current-limited power suply is your friend
[19:34:04] <cradek> p
[19:34:10] <MASEngr> He learned to check his supply after that.
[19:34:22] <MASEngr> The last guy was using 25V and a few amp.
[19:34:38] <MASEngr> He was planning to use 5V and less than 100mA.
[19:34:45] <MASEngr> Wire, wire, click, boom.
[19:34:53] <lerman> cradek: one down; two to go. I fixed the comment within a 'call' problem.
[19:35:00] <cradek> lerman: cool
[19:35:44] <MASEngr> All right, I'm going to go get a mop (figuratively speaking) then play with the scope.
[19:58:22] <tomp> haha ? mirror typing? copley's tool for tuning their drives is CME2
[22:43:10] <SWPadnos> heh - great bumper sticker: "War is not pro life" :)
[22:48:48] <SkinnYPuppY> I always liked "Guns cause violence like flies cause garbage"