#emc | Logs for 2009-09-30

[00:00:26] <Bridgeport_II> If you want to eat in Omaha, there are other places. I'll have to rely on Chris and Jeff for suggestions about Lincoln
[00:01:31] <SWPLinux> I'm sure I'll have between 0 and -1000 minutes for fun, but I'll try :)
[00:02:17] <jymm> SWPLinux: Not to worry, you're a fast eater =)
[00:02:37] <jymm> http://notalwaysright.com/why-you-never-ever-share-toothbrushes/1482
[00:03:03] <SWPLinux> that one is funny
[00:03:11] <SWPLinux> I've been reading those - they're hilarious
[00:03:30] <sed_> is there a fix for the verify function in emc2, emc1 used to show the line the error was on.
[00:03:53] <sed_> tkemc is the GUI
[00:04:48] <jymm> Yeah, There are some premo ones too, some of this crap is hard to believe ppl ar e that... that, um _______________ ! (note that is a fill-in-the-blank, as there are no words to truely describe some of these people)
[00:05:20] <SWPLinux> ok, time to go catch a plane
[00:05:27] <SWPLinux> (much easier before it starts moving)
[00:05:46] <jymm> SWPadnos: lol, have a ncie flight
[00:44:11] <cradek> in omaha, mcfosters. in lincoln, thai house
[01:13:39] <sed_> I ahve a problem with cutting arcs and circles. The Circle is not round. The Y direction is correct but the X axis is .015" oversize on an od cut. What causes this with direct read linear scales?
[01:37:11] <skunkworks> scale is correct?
[01:43:38] <cradek> seems like .015 could be mechanical too, depending on the machine
[01:55:31] <sed_> The scale is correct. This started when I upgraded from EMC1 to EMC2. Also it is not mechanical. The encoders are linear so there is no backlash in the measuring system.
[02:04:57] <sed_> I'll check back later gotta go run an errand.
[04:53:42] <turkshead> hey, does anybody have an emc2 config for a stock maxnc 10?
[09:26:03] <piasdom> g'mornin all
[09:27:27] <micges_work> hi
[10:57:57] <MarkusBec_away> MarkusBec_away is now known as MarkusBec
[12:28:31] <jepler_> jepler_ is now known as jepler
[12:42:53] <archivist> it occurred to me that it is an error to use both edges of the opto signal for threading due to it not being exactly 50% mark space so one edge has a positional error that must be ignored
[13:04:14] <cradek> but counter mode does not do that
[13:04:54] <cradek> turkshead: I think there have been many maxnc 10s: do you have a step/dir or a winding-activation one?
[13:07:36] <archivist> cradek, does it use single edge or what
[13:08:05] <cradek> encoder has a one-phase, unidirectional mode called counter. In this
[13:08:05] <cradek> mode, the phase-B input is ignored; the counts increase on each rising
[13:08:06] <cradek> edge of phase-A.
[13:08:20] <cradek> (^ man encoder)
[13:10:30] <cradek> but in the case of andypugh's encoder, which I think he said is lines drawn on the spindle somewhere he's got positional error of the same kind, but on the rising edges too
[13:10:47] <archivist> I hope andypugh gets on and makes a higher resolution encoder disk
[13:11:02] <cradek> I suspect it's more about the quality than the resolution
[13:11:22] <cradek> but I haven't seen the plot I keep asking for (!!) so I'm just guessing
[13:11:28] <cradek> so frustrating
[13:11:31] <archivist> did you read the encoder thread from him
[13:11:46] <cradek> yes
[13:12:01] <archivist> I nagged for you last night, got nowhere
[13:17:02] <cradek> he could take encoder position or encoder interpolated-position, then feed it through a ddt block, then plot the result. If it isn't pretty steady and doesn't reflect his spindle velocity well, his encoder just isn't going to work.
[13:18:11] <cradek> in theory someone could rewrite the spindle-following algorithm to work better with bad quality position feedback, but that's probably unlikely to happen
[13:20:39] <cradek> my guess at what is happening is that his pass starts at an initial wrong velocity because position difference in adjacent cycles divided by cycle time gives a terrible approximation of his velocity. that's how the pass starts. then because it starts going way too fast, it can't settle. this is why he says some passes sound worse than others. it's just luck whether you get a well- or badly-spaced set of counts for each pass's start.
[13:20:40] <archivist> he does engine management software so may just do that
[13:22:12] <cradek> that would be nice. if he can do it without breaking anything else, the change would probably be accepted.
[13:22:55] <cradek> a different scheme might use full pid. but with crap feedback, that's not going to be easy to tune either, and after much futzing you'll still want better feedback...
[13:23:36] <archivist> I agree better feedback is number one
[13:24:01] <cradek> the way the spindle's already spinning by and you have to catch up makes the whole thing challenging
[13:24:57] <cradek> 1ppr with interpolated encoder may work much better for him because you CAN get the spacing exactly right with 1ppr
[13:25:45] <cradek> bbl, want coffee
[13:48:50] <SWPadnos> jymm, http://www.newegg.com/Product/Product.aspx?Item=N82E16822145276
[13:56:16] <jymm> SWPadnos: Glad to see you made it home safely
[13:56:25] <SWPadnos> thanks
[13:56:37] <jymm> SWPadnos: how much you wan tto bet that that drive is not copatable with Thecus 4200Pro
[13:56:48] <jymm> 4100Pro
[13:57:10] <jymm> http://www.newegg.com/Product/ProductReview.aspx?Item=N82E16822102026
[13:57:55] <jymm> SWPadnos: the box I've been using for a while had its FW port die - enough to stop XP from boot with ext drives attached.
[13:58:20] <SWPadnos> hmm
[13:58:47] <SWPadnos> the only thing that looks related to drive capacity is "Capacity: 4TB or above" (on the spec sheet)
[13:58:57] <jymm> SWPadnos: VMWare Fusion doens't support FW, and OSX can't write to NTFS parttions *sigh*, so I'm kinda screwed and looking at that NAS box. $600 Raid card is just not in the budget atm.
[13:59:05] <SWPadnos> heh
[13:59:14] <SWPadnos> we had that problem on Monday
[13:59:33] <jymm> budget? or compatability?
[13:59:40] <SWPadnos> I couldn't write to HFS with a Linux machine, and the macs couldn't write to a WD drive that was formatted NTFS
[14:00:09] <SWPadnos> a little 500GB external drive, and it was NTFS formatted out of the box - kind of interesting
[14:00:22] <jymm> SWPadnos: WalMArt sells the Thecus 4100Pro for < $400 + $55 for 2yr warranty.
[14:01:05] <jymm> and no restocking fees
[14:01:12] <SWPadnos> plus tax
[14:01:25] <jymm> doens't newegg charge tax too?
[14:01:52] <SWPadnos> depends on where you live
[14:02:28] <jymm> ah
[14:02:40] <jymm> what about warranty?
[14:03:03] <SWPadnos> well, they're about to pay me $200 for a 1 year old video card that failed
[14:03:07] <SWPadnos> or send me a new one
[14:03:18] <SWPadnos> I did have to pay for return shipping though
[14:03:19] <jymm> California residents are charged sales tax
[14:03:23] <SWPadnos> ok
[14:03:24] <jymm> bastards
[14:03:53] <SWPadnos> so, the NewEgg page shows power usage with 1.5TB drives. I can't think of a limit that would appear between 1.5 and 2 TB
[14:04:20] <jymm> Newegg charges the appropriate state and/or local sales tax for any order shipping to California, New Jersey, and Tennessee.
[14:04:31] <SWPadnos> their warehouse locations
[14:04:34] <SWPadnos> oh well
[14:04:39] <jymm> yeah.
[14:06:10] <jymm> http://www.thecus.com/download.php?type=h&pid=77
[14:06:13] <jymm> SWPadnos: ^^^^^^
[14:06:30] <jymm> with beta FW
[14:06:57] <SWPadnos> well, there you go
[14:07:06] <SWPadnos> I wonder what OS it's running
[14:07:12] <jymm> linux
[14:07:21] <SWPadnos> oh. well that's handy
[14:07:28] <jymm> http://naswebsite.com/wiki/Category:Thecus_N4100PRO
[14:09:02] <EbiDK|AWAY> Anyone here on Google Wave yet and feel like sending me an invite?
[14:09:17] <jymm> SWPadnos: It supports a USB UPS, and their higher models support a Serial UPS. I *think* their might be a header on the mobo to add a DB9 out the back, just need to see if I can get support for it in the FW
[14:10:07] <jymm> EbiDK|AWAY: https://services.google.com/fb/forms/wavesignup/
[14:10:29] <EbiDK|AWAY> Yeah, I'm impatient
[14:10:49] <jymm> EbiDK|AWAY: you can't get invited by another user, by google staff only.
[14:10:54] <EbiDK|AWAY> I'm not in the forst 100K they let on there today
[14:10:58] <jymm> it's not like gmail was.
[14:11:11] <jymm> give them 21 days
[14:11:16] <EbiDK|AWAY> I read they opened that up today
[14:11:35] <jymm> then get your email addr signed u already
[14:11:40] <jymm> up
[14:11:52] <EbiDK|AWAY> Did about a month ago
[14:15:10] <jymm> SWPadnos: http://blog.backblaze.com/2009/09/01/petabytes-on-a-budget-how-to-build-cheap-cloud-storage/
[14:19:27] <jymm> SWPadnos: The 2TB hdd was on sale at frs for $179
[14:36:50] <MarkPictor> hey guys.
[14:37:20] <MarkPictor> anyone on here use jepler's gcode.ulp? I have a couple questions...
[14:37:44] <skunkworks_> hey mark@
[14:37:45] <skunkworks_> !
[14:37:58] <MarkPictor> hey skunkworks.
[14:38:24] <MarkPictor> what bit do the files start off with? the distance between paths is too wide for the engraver
[14:38:55] <MarkPictor> the first bit is not described in the file, that I can find.
[14:39:08] <skunkworks_> It starts with the clearing mill... (defaults to .0625)
[14:39:23] <MarkPictor> ok
[14:39:44] <MarkPictor> thought it looked like that, but I assumed it wouldn't use the same tool twice
[14:39:46] <skunkworks_> http://www.youtube.com/watch?v=zSgiXGELjbc
[14:39:59] <MarkPictor> thx
[14:40:25] <skunkworks_> you don't want to cut the outside of the board until the end :)
[14:40:38] <MarkPictor> ok
[14:40:53] <skunkworks_> plus there might be holes that are bigger and can be milled instead of drilled.
[14:40:59] <MarkPictor> heh, thought that video was gonna be a pcb being etched :-)
[14:41:00] <skunkworks_> (holes are done last)
[14:41:06] <skunkworks_> heh - no.
[14:41:55] <MarkPictor> ok. so fine-mill is the .01 engraver, right?
[14:42:34] <MarkPictor> err, you answered that by email
[14:42:39] <cradek> skunkworks_: that's the second time I've tried to make it through that video - still no luck
[14:43:14] <MarkPictor> ok, cool, I can give piasdom the go-ahead now
[14:43:19] <MarkPictor> thanks skunkworks
[14:45:24] <skunkworks_> cradek: too painful?
[14:45:47] <cradek> I think I just hate autotune
[14:46:16] <skunkworks_> the guy at work here says that most singers now have to use 'pitch fixers' to sound good. :)
[14:49:03] <cradek> I bet that's somewhat true, but also as music is more and more "produced" maybe we've grown less tolerant of any imperfection
[14:51:21] <cradek> I think I'm going to a "rural music" festival this weekend
[14:51:50] <cradek> they don't call them "country" anymore because people confuse that with the crap oozing out of nashville
[14:52:40] <cradek> I like seeing real people sing and play instruments
[14:57:15] <skunkworks_> yes.
[14:57:54] <skunkworks_> mother is taking mandolin lessons. :)
[14:58:25] <skunkworks_> I wish I had time to do learn an instrument. Maybe when I become rich.
[14:58:34] <skunkworks_> ;)
[15:00:43] <skunkworks_> how can steve want to average the encoder output but think 'interpolated' is bad? ;)
[15:01:16] <cradek> no idea
[15:01:38] <skunkworks_> I say tomato - you say tomato - I say potato - you say potato. (well that doen't come across very well)
[15:01:45] <cradek> haha
[15:02:00] <cradek> let's call the whole thing off
[15:02:05] <skunkworks_> ;)
[15:06:30] <skunkworks_> MarkPictor: the only thing I can think of is - I did the holes for the capasitors manually on the drill press. (only drills a small (aprox 1/16) hole.
[15:18:31] <archivist> cradek, with coarse threads index only wont/cant allow for spindle speed loss due to cut load, I know there is a delay on this lathe for the spindle control to react to the load
[15:19:00] <cradek> yep that's true
[15:19:20] <cradek> I was just saying that 1ppr might be better than a homemade encoder with the lines not even
[15:19:29] <archivist> true
[15:19:52] <archivist> but he wants to cut 4mm pitch in stainless
[15:20:14] <cradek> yikes
[15:20:56] <archivist> yup, high res encoder rules I think
[15:21:05] <cradek> brb
[15:33:02] <skunkworks_> omg - I think I should have stayed in bed.
[15:33:31] <archivist> warmer there
[15:34:08] <archivist> Ive just come inside after changing a roll bar link ready for the annual test
[15:56:26] <skunkworks_> roll bar link?
[15:59:01] <MarkPictor> MarkPictor is now known as MarkPictor-away
[16:03:59] <archivist> car anti roll bar
[16:04:24] <archivist> just went out and changed the nearside
[16:09:57] <ZeroFlex1> ?
[16:10:00] <ZeroFlex1> nearside ?
[16:10:12] <archivist> next to kerb
[16:10:23] <ZeroFlex1> not ringing a bell here
[16:10:24] <ZeroFlex1> :)
[16:10:52] <ZeroFlex1> kerb frontside house between house en street ?
[16:10:59] <archivist> lefthand side of a car in the uk
[16:11:39] <ZeroFlex1> owkeej
[16:11:49] <ZeroFlex1> learned a thing today :D
[16:12:14] <ZeroFlex1> funny, and i;m watching right now, mark evans with an mg is reborn
[18:24:45] <NewType> I got an error when I "copy" the configuration file from one computer to another. The version of EMC may be slightly different (I haven't checked, but they should be the same). What's in the configuraiton file that is computer specific? The error shows up when I select the configuration and launch... something like error ...f1.f2.f3... I can't remember.
[18:25:49] <archivist> how can we guess with something like .....
[18:25:54] <cradek> not much to go on there, for pete's sake
[18:26:15] <cradek> how do you "copy" it?
[18:27:28] <mozmck_work> If you have a mesa card on one computer and not on the other I think that would do it, or if you're using the parport and the address is different, or....
[18:29:26] <cradek> sure, and we could guess lots more reasons
[18:29:36] <cradek> knowing the error will help us guess the right one
[18:31:51] <cradek> sadly http://catb.org/~esr/faqs/smart-questions.html doesn't even cover this
[18:43:32] <jymm> lol @ cradek
[18:49:27] <skunkworks_> boy - I don't get what eu-0001 wants.
[18:51:16] <skunkworks_> he wants it to be fast and follow the path? ;)
[18:54:14] <archivist> his side tolerance is less than the end
[18:57:08] <skunkworks_> hmm - I think that is splitting hairs.
[18:57:53] <archivist> no I dont think so if you are producing a register to a spec
[18:59:17] <cradek> since the limiting factor of this kind of motion is slowing to a stop and reversing, that's so dominant that the other motion (tiny arc offset etc) is irrelevant
[19:00:19] <cradek> I think he could program it as a skinny rectangle and use plain G64 and get the best possible outcome
[19:03:25] <cradek> I bet programming tiny arcs and a skinny rectangle will give him virtually the same path and will take virtually the same time
[19:03:42] <andypugh> I wonder who you can be talking about?
[19:03:47] <cradek> there is no extra motion that will magically make it faster to reverse
[19:04:18] <andypugh> Is there a magic way to get more accelleration?
[19:04:18] <cradek> andypugh: "eu"
[19:04:32] <andypugh> (Yeah, I was being sarcastic, sorry)
[19:04:36] <cradek> sure, just edit the ini file
[19:04:42] <cradek> oh there's plenty of that going around, no worries
[19:05:10] <andypugh> I can get the machine to follow my 4mm pitch if I double the acelleration.
[19:05:32] <andypugh> Unfortunately it then drops steps mid-thread, which is not really better
[19:05:33] <cradek> we've been talking a lot about you lately. have you watched the logs?
[19:05:55] <andypugh> No, sorry. Perhaps I should.
[19:06:02] <archivist> logger_emc: bookmark
[19:06:03] <archivist> Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emc/2009-09-30.txt
[19:06:20] <archivist> the laggy bot
[19:09:15] <archivist> doesnt help that the email thread has steves problems mixed in
[19:09:16] <andypugh148> Argh! closed the wrong tab. Did I miss anything?
[19:09:45] <andypugh148> ANd you are comoletly correct about my inadequacies
[19:10:01] <archivist> your other client is still here you missed nothing :)
[19:11:08] <andypugh148> I am definitely concluding that encoder jitter is setting off problems that I don't have the accelleration to rescue.
[19:11:25] <andypugh148> And with the accell experimentally divided by 10 I get 2 second oscillations.
[19:12:01] <archivist> a single slot would fix that BUT your job in stainless at 4mm needs better control, make a better encoder wheel
[19:12:09] <andypugh148> Now, that does seem odd, it can only be the velocity that is "remembering" the problem over that many thiousands od servo threasds.
[19:13:00] <andypugh148> I would be making the encoder now, except Cromwell Tools are out of slitting saw blades, so I am having to go mail-order
[19:13:50] <archivist> * archivist eyes his stash of slitting saws
[19:14:21] <andypugh148> Incidentally, I do get a significant improvement with a faster base thread but leaving the servo thread at 1mS
[19:14:52] <SWPadnos> are you using position-interpolated?
[19:15:55] <andypugh148> However, I also still think that there should be no real need for such a responsive system if the carriage can actually wind up to the required speed in a reasonable distance (ie a few mm)
[19:16:43] <andypugh148> It needs to get to the right speed and the right place, and the current scheme is a little simplistic.
[19:16:48] <cradek> andypugh: it does seem to work well if the spindle feedback is consistent - accelerates as hard as possible up to the needed speed, then tracks.
[19:17:06] <cradek> if you get the initial speed wrong, it doesn't recover well
[19:17:36] <archivist> it has to track well to see spindle load and react
[19:17:52] <andypugh148> I wonder if it would help to get a better speed estimate all the time it is waiting for an index?
[19:17:52] <cradek> andypugh: if you can improve it (without breaking anything: the hard part) I'd be happy to review a patch
[19:19:01] <cradek> andypugh: that would fix the problem sometimes but not other times (when you happen to be right on the index to start)
[19:19:14] <andypugh148> If I can't solve it with a better encoder and (possibly) more volts, I might have a look. But C is not a language I am at all familair with (or is that C++?)
[19:19:34] <cradek> so you'd get variability that still needs to be solved some other way, like more consistent feedback
[19:19:46] <cradek> all of emc's realtime code is straight C
[19:20:06] <cradek> also all of HAL
[19:20:18] <cradek> the rest is a hodgepodge of various languages including C++
[19:20:54] <andypugh148> If I was looking at how to do it, I think I would be feeding my measured encoder signal into a Matlab model
[19:21:33] <andypugh148> But first I need a better encoder (and one that doesn't keep peeling off and giving an extra index pulse) :-)
[19:22:32] <cradek> the most obvious problem is when the index goes by, the position starts counting up but you're still stopped and have a lot of accelerating to do. so currently the algorithm is to accelerate and track a new position that's behind the original, but that it can get to.
[19:23:30] <cradek> a different algorithm might use the knowledge that the spindle counts in a predictable way and you don't HAVE to follow the initial reset - it doesn't matter which turn you follow
[19:23:41] <SWPadnos> maybe it should use 3 indexes (possibly 4, if it's on an index at the beginning)
[19:24:28] <andypugh148> Does it try to guess the finish velocity? It looks to only really consider position, whereas I can see trying to blend to a "move just like this one" might work better.
[19:24:28] <SWPadnos> the first pair of indexes would give a baseline speed, so some acceleration could be applied before the next index is expected
[19:24:39] <archivist> you know the revs and acceleration rate calc how many are needed
[19:25:02] <cradek> not sure what you're asking
[19:25:25] <andypugh148> I need to sit down with a pen and paper and go through the code
[19:25:26] <cradek> SWPadnos: yeah it seems like anticipating the position is needed
[19:25:57] <cradek> it is definitely a problem in the position realm - often people get confused and think it's a question of velocity
[19:26:01] <andypugh148> But my feeling is that it needs to work in velocity space as well as position space.
[19:26:14] <archivist> you could also define how far off the stock so you can get to speed
[19:26:17] <SWPadnos> "whichever is more accurate"
[19:26:24] <andypugh148> (People like me, it seems :-)
[19:26:31] <cradek> heh
[19:26:53] <cradek> well if you match position you automatically match velocity too
[19:27:15] <cradek> but I think we all see why matching just velocity is not sufficient
[19:27:49] <andypugh148> True, but does it try to match position with zero velocity? (I am assuming not)
[19:28:01] <cradek> the current algorithm gives great (perfect?) performance on my lathe but I have lots of encoder resolution
[19:28:04] <andypugh148> Because that is too obviously silly
[19:28:11] <cradek> that's why I thought it was solved "well enough"
[19:28:32] <cradek> sorry what do you mean with zero velocity?
[19:28:58] <andypugh148> It isn't trying to stop at the end of the queued move?
[19:29:17] <andypugh148> It does look like it might be trying to stop on an overshoot
[19:29:19] <cradek> if that is the last move, yes definitely it will decelerate and stop
[19:29:40] <cradek> if it is not the last move, it blends the motion into the next one without stopping
[19:29:59] <andypugh148> But I have been too busy to follow the code through to where Maxvel and Accell etc go to
[19:30:00] <cradek> g76 always has the equivalent of g33 "over" and then g0 "out"
[19:30:34] <cradek> which gives a very nice thread end
[19:30:54] <andypugh148> Does it need a runout groove? I suspect not, which is nice
[19:31:02] <cradek> nope
[19:31:20] <cradek> (if it could not do that blend it would definitely need one)
[19:31:27] <andypugh148> Anyway, you were wanting a Halscope trace of my encoder? (if it deserves to be called an encoder)
[19:31:47] <cradek> I wonder what your position and interpolated-position show *on each servo cycle*
[19:32:13] <cradek> I bet interpolated- is still not smooth, if your encoder line spacing is variable, and I am mostly curious how bad it looks
[19:32:52] <andypugh148> So, sample at 500uS, plot at 10mS per division and pastebin the raw data?
[19:32:58] <tomp> tomp is now known as tom3p
[19:33:34] <cradek> just the picture is easier to understand
[19:33:57] <andypugh148> It looks perfect on the picture, I think. Let me do both
[19:34:05] <andypugh148> (I will be back)
[19:34:14] <cradek> the only pictures I've seen do not have distinguishable individual servo cycles
[19:34:26] <cradek> ok
[19:39:01] <cradek> about your difficult large thread: do you know the trick for avoiding doing full depth at once? you can do a pass to half depth offset by (mumble) in Z that clears the outer half of the full depth thread
[19:39:28] <cradek> offset by half the pitch I bet
[19:40:24] <cradek> I can't find a picture online...
[19:40:42] <archivist> hmm does g76 know abot acme and buttress thread
[19:40:59] <cradek> you can do compound infeed at any angle you like
[19:41:20] <cradek> assuming that's what you mean?
[19:41:34] <archivist> acme is a square thread
[19:41:55] <cradek> it certainly is not!
[19:42:27] <archivist> trus 14 deg, I mean a real square thread
[19:42:53] <cradek> you could do real square thread with G76 Q0
[19:42:55] <BJT-Work> this one: http://en.wikipedia.org/wiki/Acme_thread_form
[19:43:04] <cradek> acme would be Q14.5 or whatever the number is
[19:43:18] <cradek> well you can cut any thread with Q0 but you usually don't want to...
[19:44:07] <archivist> there is a left then right move recommended as well
[19:44:35] <cradek> I have never cut a square thread - hope I don't ever have to
[19:45:11] <archivist> I have, steam valve stem about 12" of thread
[19:45:25] <archivist> 1/2" pitch
[19:45:41] <cradek> square tool tilted to the helix angle, straight in?
[19:46:04] <cradek> or do you use several tools ... somehow
[19:46:50] <archivist> took a while, did it on my southbend, do a cut fiddle, rinse repeat
[19:48:01] <archivist> resharpen tool mid job as well
[20:12:53] <andypugh148> Bother! It isn't working at all now, I can't even home. I thought the leadscrew was seized, but I can't home with the motor disconnected, I have no idea what has happened
[20:13:33] <andypugh148> Though I guess that doesn't stop me getting the halscope, as the motors don't actually have to run..
[20:15:20] <archivist> I had a wire fall off the other day
[20:20:28] <jymm> duct tape
[20:22:46] <cradek> andypugh: yuck, what happened to it?
[20:32:44] <NewType> yeah, I know my description was a bit vauge.
[20:32:52] <NewType> this is what I did:
[20:33:14] <NewType> I copied the "configs" folder from one computer to another.
[20:33:46] <NewType> after launching EMC, I select the newly transfered machine configuration in the tree.
[20:34:20] <NewType> then I have this "Application Error" that begins with :
[20:34:20] <NewType> .main.f1.f2.f3.tree selection set: Cannot select unknown node
[20:35:14] <cradek> that seems unrelated to your configuration. can you run sim/axis?
[20:35:42] <NewType> yep.
[20:36:01] <NewType> I can start a new EMC session, create new configuraiotn, and EMC/axis will run fun
[20:36:04] <NewType> *fine
[20:36:27] <NewType> I am actually merging two control computer into one.
[20:36:42] <NewType> so I am copying the configuration over from one computer to another.
[20:37:09] <NewType> I didn't think the configuration has special ties to a specific machine... that was puzzling.
[20:37:34] <archivist> internal addresses of ports can be different
[20:37:47] <NewType> TRUE! I forgot about that!
[20:39:44] <NewType> is that in *.hal?
[20:41:32] <NewType> let me quickly check...
[20:43:55] <_Poincare> _Poincare is now known as Poincare
[20:50:02] <andypugh148> Well, I don't know what has happened, but my Z-axis is down to less than half speed. I can't imagine why, though the motor is making a ticking noise.
[20:55:27] <andypugh> Annnoyingly there seems to be some sort of conflict betwixt this embedded client and pastebin
[20:55:41] <archivist> embedded client?
[20:56:06] <archivist> I use xchat
[20:57:24] <andypugh> The one that you can click on the linuxcnc "community" page
[20:57:37] <andypugh> Anyway, for cradek: http://imagebin.ca/view/ud8Tect.html
[20:57:53] <archivist> I never use it, its less usable
[20:58:04] <andypugh> Though suddenly the encoder problem is not my main worry
[21:04:39] <Megazone> nope. all the working config and the transfered config are using the same paralle port address.
[21:04:44] <Megazone> woops.
[21:05:37] <Megazone> Megazone is now known as NewType
[21:05:47] <NewType> there.
[21:13:33] <NewType> still nothing. I checked the copied files for permission problem, and I changed them again. the copy command didn't set the permission correctly.
[21:18:17] <andypugh_> Proper client now, it's got an insertion point cursor and everything!
[21:18:55] <archivist> bout time too :)
[21:19:32] <jymm> what archivist said
[21:19:41] <andypugh_> Tomorrow I think I will either be dismantling a stepper motor or trying to clean a ballnut.
[21:20:22] <jymm> You do that, and you'll never have a stepper motor again.
[21:20:23] <andypugh_> Making the encoder is looking unlikely
[21:20:30] <andypugh_> ? Nothing magic in a stepper motor
[21:20:40] <archivist> I have to get the mundano through its mot tomorrow
[21:23:02] <andypugh_> Unless there is something magic in a stepper motor, and I let it out when I had it apart last time to machine the end-cover to take a conduit connector, of course.
[21:24:03] <andypugh_> I rate my chances of cleaning and reassembling the ballnut as very low indeed, though.
[21:28:30] <jymm> I can't find the article, but this place rebuilds motors and this is what they say: "Why do we re-magnetize all motors?
[21:28:34] <jymm> It is important to understand that once a servomotor, stepper motor or tachometer has been taken apart for repair, it will usually not work properly again until the magnetism is restored to its proper level. "
[21:28:56] <jymm> http://www.servorepair.com/why_choose.htm
[21:31:08] <andypugh_> Well, it is too late now, as I have already had it apart once, and I suspect that repair would cost more than replacement.
[21:31:17] <DaViruz> i have taken apart both stepper motors and servo motors without any changes in performance (that i can detect)
[21:33:12] <archivist> uncouple from screw and test first
[21:34:00] <DaViruz> although i can imagine that letting the rotor impact the stator at the force of the permanent magnets can be a bad thing
[21:34:07] <andypugh_> That is what is worrying me, it is maxing out at half the old coupled speed, uncoupled.
[21:34:33] <DaViruz> but i can't think of anything else that would change the permanent magnetism in the motor?
[21:35:29] <andypugh_> I did up the current limit from 2.1 to 2.5 A in search of more accelleration, but it didn't even get warm so I doubt it it actually burnt out.
[21:36:52] <archivist> cold steppers is a sign that you can drive them harder
[21:37:31] <skunkworks_> what voltage?
[21:37:32] <andypugh_> I need to check that I didn't disturb anything else while switching current limit, I suppose.
[21:37:41] <andypugh_> 27V
[21:37:47] <skunkworks_> what drives?
[21:38:00] <andypugh_> Motion Control Products
[21:38:19] <skunkworks_> you will probably get a lot better performance if you can raise the voltage.
[21:38:21] <jymm> *typically* 20x to 25x rated voltage
[21:38:32] <andypugh_> http://www.motioncontrolproducts.com/drives/msd325-microstepping-drive.php?cat=2
[21:38:54] <skunkworks_> eww - 32c
[21:38:55] <andypugh_> Yes, that was actually why popped on here for a few minutes a few hours ago
[21:38:55] <skunkworks_> 32v max
[21:39:32] <jymm> ew, that's a glorified xylotex
[21:40:21] <skunkworks_> what current are the steppers rated at?
[21:40:27] <jymm> No, I take that back... its LESS THAN a Xylotex, max current: 2.5A
[21:40:43] <andypugh_> The motors can be wired parallel or series, I have them series/2.5A/5.5V instead of parallel/4.2A/2.73V
[21:42:00] <andypugh_> The motor spec-sheet shows 3x the torque at 3000pps for parallel than series, but I am guessing that is at rated voltage, and I can get the same effect with more volts?
[21:42:02] <jymm> $60 USD - OUCH!
[21:42:09] <jymm> single axis?
[21:42:28] <andypugh_> I have 4 axes
[21:42:54] <jymm> andypugh: do you have 4 of tose things yu linked to?
[21:42:57] <jymm> those
[21:43:03] <andypugh_> Yes
[21:43:24] <jymm> 4x $60 = $240 USD <------ OUCH!!!!!!!
[21:43:27] <andypugh_> They seemed to get decent reviews on CNCzone
[21:44:38] <jymm> For $40 more you get a WHOLE LOT MORE... http://www.geckodrive.com/product.aspx?c=3&i=14469
[21:45:11] <skunkworks_> bbl
[21:46:58] <andypugh_> And were a definite upgrade from the eBay special triple-TA8435H from china that I kept blowing up.
[21:48:49] <andypugh_> I don't think that the G540 existed when I was shopping (about this time last year). I looked at other Geckos but by the time shipping was factored in they looked very expensive, and on paper were no great advantage for the motors I already had.
[21:49:45] <archivist> some geckos have had odd problems
[21:50:15] <archivist> I think most drives have foibles
[21:50:17] <andypugh_> Yeah, I was also swayed by having the supplier in the same country. It's a long swim to Gecko-land
[21:51:46] <archivist> I get mine from arc eurotrade but at least I can nag him
[21:53:43] <andypugh_> Anyway, is the difference between series and parallel winding purely down to inductance and easily overcome by More Volts? It seems likely to me. It is 2.1A per winding regardless of wiring scheme, after all?
[21:55:52] <andypugh_> (Bah! Cradek lambasts me behind my back for not producing the plot he wanted, then disappears as soon as I create it)
[21:56:22] <archivist> 30/5 is only 6 and 30/2.5 is 12 a far better ratio
[21:56:45] <archivist> *typically* 20x to 25x rated voltage
[21:58:00] <alex_joni> andypugh_: so where is it?
[21:58:47] <andypugh_> http://imagebin.ca/view/ud8Tect.html
[22:00:02] <alex_joni> I think cradek left work about 1h ago
[22:00:20] <alex_joni> he'll probably be online in a while
[22:03:06] <alex_joni> andypugh_: any more results from tests?
[22:04:29] <andypugh_> No, I seem to have broken something trying to squeeze more accelleration out of it. Even uncoupled the motor is only managing half the speed it was previously when connected.
[22:05:32] <archivist> those dip switches to set the current, make sure they are correct (not inverted)
[22:06:22] <andypugh_> They are right, I am confident of that, and now back where they were. I might have nudged a microstepping one, possibly.
[22:06:49] <andypugh_> I will plug that axis into a different amp, and that amp into a different axis and see what I get tomorrow.
[22:07:49] <archivist> I see similarity on that image to mine
[22:08:20] <archivist_emc> http://imagebin.org/65943
[22:12:29] <andypugh_> Your encoder consistency actually looks worse than mine.
[22:12:57] <andypugh_> You should try a bit of laserprint wrapped round the spindle :-)
[22:13:21] <archivist> this is the original factory fitting
[22:14:17] <andypugh_> I know. Is it the original factory detectors?
[22:14:47] <archivist> yes
[22:14:57] <andypugh_> I run mine through comparators on the board to boost level and sharpen the edges, does your have any>
[22:15:11] <archivist> directly connected no opto
[22:16:40] <archivist> I think mine is just pullups
[22:17:26] <andypugh_> Might help to run them through comparators as buffers. I have several more than I need.
[22:18:10] <skunkworks> increasing the base period wouldn't help?
[22:18:27] <andypugh_> Decreasing you mean?
[22:18:28] <archivist> I use a ULN2003 as a converter the other direction to drive the Parker digiplan stepper driver (12v)
[22:18:48] <skunkworks> heh - yes
[22:20:18] <andypugh_> What are your thread timings . Archivist?
[22:22:05] <andypugh_> I found today that keeping the 1mS servo thread but dropping the base thread from 100uS (stepconf chosen) to 20uS helped with the G76 issue and, oddly, seemed to allow higher axis acelleration too.
[22:23:41] <archivist_emc> my ini http://pastebin.ca/1586782
[22:24:41] <andypugh_> Aha! still on the stepconf defaults then. Try BASE_PERIOD = 25000
[22:24:59] <archivist_emc> my hal http://pastebin.ca/1586784
[22:25:28] <andypugh_> I am not sure why Stepconf asks what the latency and timings are, it never seems to choose anything other than 100,000 and 1,000,0000
[22:30:50] <Guest267> whoami
[22:30:53] <Guest267> heh
[22:31:00] <Guest267> Guest267 is now known as skunkworks
[22:31:31] <andypugh_> Those arc-eurotrade 6.5Nm steppers look like monsters. http://www.arceurotrade.co.uk/Catalogue/Stepper-Motors/Stepper-Motors
[22:33:47] <andypugh_> Ah, just noticed, 86mm is Nema 34?
[22:34:25] <BJT_Shop> dang Coco don't hang around long after the cheese is finished
[22:37:19] <archivist> andypugh, mill has 180 nm on xy and 220 on z and b a is some junk
[22:37:20] <andypugh_> <Digression> Has anyone found a handy source of spiral spring leadscrew covers?
[22:37:48] <archivist> a complete denford :)
[22:37:58] <andypugh_> 180Nm steppers? Whoa!
[22:38:30] <archivist> erm 180ncm
[22:38:39] <archivist> not a meter
[22:38:47] <andypugh_> Ah, the one I was looking at is 650Ncm
[22:39:10] <andypugh_> Mine are 300Ncm, 3Nm
[22:39:13] <archivist> they are the ones on that page that I have
[22:39:18] <andypugh_> And can't get synch to a 4mm lead...
[22:40:50] <archivist> Im rerunning at 4mm
[22:42:41] <andypugh_> I am losing my marbles, just went to see if my dinner was done, and there it was stood on top of the oven, not in it.
[22:42:56] <archivist> about 25rpm spindle so the carriage can follow at around 30 ipm
[22:43:26] <archivist> I should go home now
[23:25:21] <andypugh_> Is cradek there at the moment?
[23:26:49] <skunkworks> he is there... but I don't know if he is here..
[23:26:53] <skunkworks> ;)
[23:26:55] <Mire> Hello everyone. Today I'm battling with PID tuning. I am getting following errors on jogging at modest speeds.
[23:28:09] <skunkworks> how are you tuning?
[23:28:13] <Mire> I, too, am hopin cradek is around because I essentially copied his HNC config and I have the same machine. But I've never seen it work it's best, so I won't know when I've won.
[23:30:27] <Mire> In particular, I'm wondering how fast I should be able to jog w/o following errors. 30 in/min seems slow for this monster. What else besides PID parameters will affect this?
[23:32:13] <Mire> Tuning by the simplistic method... zero I and D, bump P till it rings, bump I till it stops...
[23:33:00] <Mire> but it is on continuous jogs that it seems to be trouble, sometimes at start, sometimes at stop.
[23:34:45] <SWPadnos> I think you're supposed to back off P a little once you get oscillation
[23:35:16] <Mire> I began with cradek's values. buzz, fear, estop. I then drastically cut the P parameter and things started to work. But the following errors appeared as I bumped up the P term, well before oscillations set in again, and far below the value I had began with.
[23:35:32] <Mire> Can a P value too low induce overshoots?
[23:35:34] <SWPadnos> do you have the same motors, drives, and feedback scales?
[23:36:09] <cradek> the amp tuning is probably totally different
[23:36:33] <Mire> yyyes. identical machine, factory servos and drives and synchros.
[23:37:01] <cradek> Mire: bump *D* until it stops, not I
[23:37:07] <Mire> hi chris. yup, very different so far. but following errors under 30 ipm?
[23:37:20] <cradek> totally mistuned then
[23:38:04] <cradek> you may very well end up with I=0 for velocity mode (and tweak the amp offset control to remove steady state error)
[23:38:33] <cradek> I think the original control did 250ipm rapids - you should be able to get good following at that speed
[23:38:47] <cradek> looks like I have my error trip set at .0015
[23:39:12] <cradek> you will have to set it MUCH higher (.05? .1?) for tuning or it will trip all the time
[23:40:09] <cradek> jogs have higher acceleration than (most) mdi/program motion so they will be the worst at the start/end of the move
[23:41:18] <andypugh_> Low P can cause overshoots if I is high. You get "I-term windup". ie I increases all the time that P is failing to make the setpoint, then when you finally get to the setpoint, the I term is still telling it to move further.
[23:41:49] <cradek> yes andypugh_ is right. I should be the very last thing you set, if you need it, once your following is very good otherwise.
[23:41:51] <alex_joni> cradek: did you see andypugh_ halscope screenshot?
[23:42:04] <cradek> nope, looking
[23:42:13] <alex_joni> http://imagebin.ca/view/ud8Tect.html
[23:42:13] <Mire> that was a good clue. i suspected that it mite be laced up too tight!
[23:42:56] <cradek> haha, nothing is behind one's back on irc
[23:43:15] <alex_joni> heh
[23:43:24] <cradek> on the contrary I expect people to magically know what I say, even if they're not there at the time!
[23:43:34] <alex_joni> * alex_joni is off to bed..
[23:43:37] <alex_joni> good night all
[23:43:42] <cradek> goodnight alex_joni
[23:43:42] <skunkworks> night alex
[23:43:50] <alex_joni> oh, and behave.. I'll be watching :D
[23:43:55] <cradek> heh
[23:43:57] <cradek> who me?
[23:44:17] <alex_joni> * alex_joni looks behind cradek's back
[23:44:41] <cradek> andypugh_: that plot is not as bad as I expected after seeing jepler's (purposefully) bad results - the interp output is pretty decent in your case
[23:45:09] <alex_joni> even not interp isn't that bad
[23:45:16] <andypugh_> Yes, I admit to some surprise.
[23:45:33] <alex_joni> andypugh_: maybe you can keep trying capturing a bad one?
[23:45:34] <cradek> well for this naive? algorithm the non-interp is bad bad bad
[23:45:44] <andypugh_> I might re-christen my printer my "CNC pigment application device"
[23:45:54] <MarkusBec> MarkusBec is now known as MarkusBec_away
[23:46:08] <Mire> One last thing cradek: I get .002" backlash in Z. for this machine is that a lot?
[23:46:21] <cradek> the initial velocity it sees is one of those steps which represents 10x the real velocity
[23:47:09] <andypugh_> I have no idea if that was a good or bad run, the z-axis is currently broken in some way yet to be determined.
[23:47:15] <cradek> Mire: yes that means your ballscrew is as bad as mine was - if the backlash is the same everywhere, the wear is in the nut or balls, and you can put in bigger balls to fix it
[23:47:43] <cradek> Mire: if X is good (mine is perfect) you are lucky because that's the really important one IMO
[23:48:48] <cradek> Mire: the nuts have two circuits for preload, but is non-adjustable. so the preload comes purely from very careful ball sizing
[23:49:06] <cradek> well the Z nut anyway - I have never touched or seen the X nut
[23:49:12] <andypugh_> How hard is it to repopulate a ballnut?
[23:49:21] <Mire> Yup that's what I get. Must be a result of usage patterns. Z's a bit growly as well. Will it be pricey to fix?
[23:49:46] <cradek> was it growly on the old control?
[23:50:30] <cradek> andypugh_: I have done two - neither was particularly hard but it takes much patience and sometimes many tries (ball sizes)
[23:51:02] <cradek> of course I'm sure getting to it is often the hard part
[23:51:58] <cradek> Mire: my Z nut housing was full of rusty water stuff where there should have been oil. I bet yours is too. You're supposed to squirt way oil in there every time you use it. I bet nobody does.
[23:52:10] <Mire> The noisy thing had four loud fans and about seven buzzy xfmrs. The motor ran at all times when the controller was powered. I ran off a 3phase converter made from an old motor. The pneumatics leaked like a sieve.
[23:52:18] <cradek> I suspect the lack of oil is what messes it up, not wear
[23:52:26] <Mire> I'm hearing new sounds now:)
[23:52:28] <cradek> haha
[23:53:06] <cradek> try driving it around with a 1.5v battery for velocity command. if it growls, you know it is the velocity loop oscillating. if so it may be mechanical or electrical or both.
[23:53:08] <andypugh_> I ask because I might have contaminated mine. I only have wiper seals and have not figured out how to protect the leadscrew yet. (looking for a source of spring-coil covers (cheap)). You can turn the handle a quarter turn each way reasonably easily, then it stiffens, presumably as the balls swap direction then start pushing each other through the return circuit.
[23:54:30] <andypugh_> However, I only found that looking for reasons that my axis performance had suddenly degraded, and that looks to be motor, drive or software.
[23:54:31] <cradek> andypugh_: they sure don't like it if they don't stay clean. the balls have no choice but to crunch up the goop since they can't push it out of the way...
[23:55:17] <cradek> I have noticed that before (a bit of a turn is easy, then gets harder) but I'm not sure I understand the cause. maybe your circuits are not totally full?
[23:55:26] <andypugh_> Aye, I know. Covers were on the to-do list, but the lathe is still making itself, and that job awaits thought.
[23:55:49] <andypugh_> Indeed, it might well be completely typical.
[23:56:00] <cradek> iirc, jmkasunich was looking for those coily spring things and was shocked at their cost
[23:56:28] <andypugh_> I can't imagine turnings getting through the wipers, and I don't make much dust.
[23:56:38] <cradek> that's good then, probably fine
[23:56:53] <andypugh_> I was thinking of buying some stainless shim and making my own spring-covers.
[23:57:01] <cradek> if there's a hole in the nut for oiling, you could run a bunch of oil through there, cover everything, and forget about it
[23:57:30] <cradek> fun, I'd love to hear how that turns out, assuming you don't bleed all over the project
[23:58:04] <Mire> Mine sounds okay when moving faster. Z servoamp will buzz in differing ways after .001 increments.
[23:58:19] <cradek> come to think of it, shim stock does come in 2-3" or so wide rolls. might just work.
[23:58:47] <cradek> Mire: it's easy to see/feel a tiny oscillation if you put an allen wrench in the screw
[23:59:06] <cradek> on mine I could even tap the wrench and set it buzzing
[23:59:31] <skunkworks> and fling it across the room ;)
[23:59:34] <cradek> fascinating that you seem to have exactly the same problems I did
[23:59:35] <andypugh_> Hmm, I have just found the quote from M Buttkereit Ltd. Open length 800mm, closed length 800mm £43.10 each. I don't know why I am messing about really.