#linuxcnc-devel | Logs for 2017-02-04

[05:32:30] <jthornton> pcw_mesa: is there a 7i92 MX4660x2D bit file for the Leadshine MX4660? seems the only difference for P2 from the MX3660 is the addition of a 4th axis like the G540
[07:57:30] <JT-Shop> I see in a note that you said you have the MX4660x2D.bit but http://freeby.mesanet.com/7i92_MX4660x2D.bit gives me a 404
[07:57:59] <JT-Shop> The requested URL /7i92_MX4660x2D.bit was not found on this server.
[09:56:06] <ASRock_pc> http://www.snopes.com/mexico-guatemala-border/
[09:56:17] <ASRock_pc> ignore that please
[09:56:20] <ASRock_pc> wrong channel
[10:25:21] <jepler> you can't go wrong with a snopes link
[12:12:36] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15jepler commented on issue #228: @SebKuzminsky I'd just take the whole text of that rule from master branch into 2.7 rather than trying to cherry-pick -- it looks like it went through a couple of revisions due to the docker bug that an intermediate version was triggering. In particular, I'm surprised to see any checks of IS_POWERPC left in that Makefile rule. 02https://github.com/LinuxCNC/linuxcnc/issues/228#issuecommen
[12:26:54] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15jepler commented on issue #225: It looks like we have to do this with several steps:... 02https://github.com/LinuxCNC/linuxcnc/issues/225#issuecomment-277463079
[12:30:07] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15jepler commented on issue #225: It looks like we have to do this with several steps:... 02https://github.com/LinuxCNC/linuxcnc/issues/225#issuecomment-277463079
[13:04:40] <JT-Shop> pcw_home: do you have a 7i92_MX4660x2D.bit
[13:14:50] <pcw_home> yeah its on our server
[13:14:51] <pcw_home> is it not in 7i92.zip?
[13:20:31] <pcw_home> in any case:
[13:20:32] <pcw_home> http://freeby.mesanet.com/7i92_MX4660x2D.bit
[13:22:21] <JT-Shop> let me double check, I know 7i92_MX3660x2D.bit is there
[13:23:14] <JT-Shop> thanks
[13:25:34] <JT-Shop> ah yea it's in 7i92.zip, looks like I needed to get a fresh copy
[13:50:27] <skunkworks> would there be a fancy way to lower the frequency of one of the pwm-gens on a mesa card?
[13:52:03] <skunkworks> darn.. scared him away
[13:52:04] <skunkworks> ;)
[13:53:31] <jepler> skunkworks: the docs claim that each pwmgen has its own pwm_frequency pin, but I didn't verify that's correct
[13:53:54] <jepler> oh wait just below that it says it applies to all instances
[13:54:08] <skunkworks> right - I see a pwm and a pdm frequency
[13:54:21] <skunkworks> but not for each
[13:54:26] <skunkworks> jepler, welcome back!
[13:54:30] <jepler> if you need two frequencies, can you get away with making all the things at frequency 1 PWM and all the things at frequency 2 PDM ?
[13:54:34] <jepler> hi skunkworks and thanks
[13:54:45] <skunkworks> would that work? that would be awesome
[13:56:44] <skunkworks> let me look
[13:57:33] <jepler> so yeah looking at pcw's "regmap" file from hostmot2 source, there's one register to control the frequency of ALL pwms, and another to control the frequency of all PDMs
[14:01:21] <skunkworks> that worked!
[14:02:05] <skunkworks> pdm at 10khz seems to actually run this 0-10v output pretty linearly
[14:03:36] <skunkworks> pcw_home, we got it.. ;)
[14:07:51] <skunkworks> within a couple 10ths from 0 to 10
[14:07:53] <jepler> hmm maybe be more careful with your shipping containers http://www.arabiansupplychain.com/pictures/box.jpg
[14:09:03] <jepler> I guess it ran aground and spilled a bunch of diesel and containers https://en.wikipedia.org/wiki/MV_Rena
[14:42:45] <JT-Shop> skunkworks: what size stepper is in the Compact 5 PC lathes?
[14:58:51] <jepler> this ebay listing suggests NEMA 23 http://www.ebay.com/itm/Emco-Compact-5-CNC-Stepper-Motor-NEMA-23-and-cable-/302197702806
[15:01:48] <JT-Shop> thanks
[15:02:08] <JT-Shop> ebay is useful after all :)
[15:26:24] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/external_offsets bacdd18 06linuxcnc 10(23 files in 9 dirs) External Axis Offset hal pins (v27) * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=bacdd18
[15:29:12] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15njh commented on issue #225: Thank you very much @jepler, was starting to worry that it had all been wasted effort 😨 ... 02https://github.com/LinuxCNC/linuxcnc/issues/225#issuecomment-277477606
[15:42:34] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15jepler commented on issue #225: hm I'm not seeing the transfer of simple-gcode-generators from @njh to @jepler yet. did you complete all 5 steps at https://help.github.com/articles/transferring-a-repository-owned-by-your-personal-account/ or did you run into trouble? 02https://github.com/LinuxCNC/linuxcnc/issues/225#issuecomment-277478349
[15:51:50] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15jepler commented on issue #225: aha found it (the notification goes to email only, apparently) 02https://github.com/LinuxCNC/linuxcnc/issues/225#issuecomment-277478906
[15:53:08] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15jepler commented on issue #225: It looks like we have to do this with several steps:... 02https://github.com/LinuxCNC/linuxcnc/issues/225#issuecomment-277463079
[15:54:38] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15jepler commented on issue #225: https://github.com/LinuxCNC/simple-gcode-generators... 02https://github.com/LinuxCNC/linuxcnc/issues/225#issuecomment-277479077
[16:12:55] <jepler> http://wafflesatnoon.com/wp-content/uploads/2013/04/math-quiz2.jpg awesome jpeg artifacts, the only trick they missed was to misspel "genuises"
[16:49:02] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/external_offsets 836a7a5 06linuxcnc 10(23 files in 9 dirs) External Axis Offset hal pins (v27) * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=836a7a5
[16:49:53] <dgarr> external offsets getting close (@commit 836a7a5) https://youtu.be/-7hjcqZkHK8
[16:54:58] <pcw_home> Wow, that looks great!
[16:56:13] <JT-Shop> sweet!
[16:56:37] <dgarr> it hasn't been easy
[16:59:16] <JT-Shop> do you think it's ready to be tested on my plasma?
[17:02:48] <dgarr> some behavior may not be intuitive so exercise the provided sim config first would be a good idea.
[17:04:10] <dgarr> i have tested on hardware -- one important concept is to have no movement at power off or power on if offsets are applied -- hence they are reset at turn on
[17:04:30] <dgarr> the asciidoc documentation (readable) is http://git.linuxcnc.org/gitweb?p=linuxcnc.git;a=blob_plain;f=docs/src/motion/external-offsets.txt;hb=refs/heads/dgarr/external_offsets
[17:06:41] <JT-Shop> I've been pulling each time you commit but still need to change the plasma from Ubuntu to Linux Mint 17.3 so I can test it
[17:07:12] <JT-Shop> it's stepper driven with a floating head so you can't hurt it :)
[17:12:49] <dgarr> i haven't found any show-stoppers, test reports welcome -- "No battle plan survives contact with the enemy"
[18:38:22] <skunkworks> JT-Shop, but they are not hybrid steppers (no maganet in the rotor)
[18:38:28] <skunkworks> so they are pretty old school
[18:41:21] <skunkworks> dgarr, how does it handle accel/vel constraints?
[18:41:45] <jepler> skunkworks: I think I saw in the docs that you specify how much of the constraints the TP gets to use, and how much the external offset gets to use
[18:41:56] <jepler> but that may have been a few iterations ago
[18:42:15] <jepler> [AXIS_L]OFFSET_AV_RATIO = value (controls accel/vel for external offsets)
[18:43:22] <skunkworks> ah - awesome
[18:45:35] <jepler> the good news: it looks very likely that debian squeeze will enter the hard freeze with a -rt kernel
[18:45:48] <jepler> the bad news: I tried the 4.9-rt kernel in jessie-backports and it didn't work with my hardware
[18:46:12] <jepler> https://packages.debian.org/stretch/linux-image-4.9.0-1-rt-amd64
[18:46:57] <jepler> (I tried it back in December and I don't remember exactly what was wrong; I think that there were problems with nvidia video)
[18:50:50] <pcw_home> I've had mixed results with 4.9.6-RT4 (terrible latency = 700 usec on Core Duo, hang on boot on I5 )
[18:52:23] <skunkworks> I have been runnig 4.8.6-rt5 on this laptop and it seems ok
[18:53:33] <pcw_home> >4.1 works on some hardware but its really hit&miss
[18:54:01] <jepler> yeah
[18:54:17] <jepler> there don't seem to be one-size-fits-all solutions right now which really stinks
[18:54:25] <jepler> (even ignoring how -rt is bad for people with dumb parports)
[18:55:04] <skunkworks> jepler, I have been running the emco lathe on uspace..
[18:55:12] <skunkworks> 50us base period
[18:55:14] <skunkworks> no issues
[18:55:31] <skunkworks> even when it is up for days
[18:55:42] <skunkworks> but that is probably running a 3.something rt
[18:56:11] <skunkworks> 3.18.9
[18:56:16] <skunkworks> usually up for days
[18:56:18] <andypugh> But us RTAI even an option any more? What’s the vintage of the last supported kernel
[18:56:32] <dgarr> skunkworks: http://buildbot.linuxcnc.org/doc/scratch/v2.8.0~pre1~dgarr-external-offsets~836a7a5/html/motion/external-offsets.html
[18:57:13] <jepler> andypugh: I don't even know, tbh
[18:58:06] <skunkworks> dgarr, awesome work. Sorry I have not had time to play with it. trying to get a lathe and edm going
[19:02:10] <jepler> same with xenomai, viz. kernel support. you can get decent (rtai-like) latencies from it in master branch with uspace, but it looks like when I played with it I used the 3.18.x kernel as the base. http://article.gmane.org/gmane.linux.distributions.emc.devel/16447
[19:03:02] <jepler> actually they've got kernel 4.1, that seems to be the latest https://xenomai.org/downloads/ipipe/v4.x/x86/
[19:11:24] <andypugh> A chap on the forum is insisting that his Fanuc drives use 6 PWM channels. 3 for FWD and 3 for REV. Does that make any sense at all?
[19:11:55] <pcw_home> AFAIK the are just top/bottom transistors
[19:12:11] <pcw_home> they are
[19:12:42] <andypugh> Yes, direct drive to the 6 transistors would make perfect sense.
[19:12:52] <andypugh> But what he says doesn’t seem to.
[19:13:53] <andypugh> TBH his friend who has told him this might indeed have been a service tech, but that doesn’t mean he needed to know the actual implementation details of the drives.
[19:15:14] <skunkworks> what would be better to run a spindle? current mode or duty cycle mode? (dc brushed)
[19:15:51] <pcw_home> PWM mode is probably easier to tune
[19:18:16] <skunkworks> I was thinking that - just from the servos are pwm - and they tuned suprisingly easy
[19:19:11] <skunkworks> This amp with take +/-10 in - but set to voltage
[19:20:22] <skunkworks> pdm seems to work with this cheap bob the best. It is +/- .01 volt from 0 to 10 at a pdm of 10khz
[19:20:42] <pcw_home> I have a couple of the PWM FANUC drives I can trace them out but a a quick glance they seem very simple
[19:20:44] <pcw_home> Direct PWM control of the transistors (PWM signals are bi-directional and output fault status when in fault states)
[19:20:46] <pcw_home> They also have +-4.4V analog feedback of 2 phase currents (so expect the controller to do FOC)
[19:26:53] <skunkworks> zlog
[19:56:22] <skunkworks> the one good thing about the -rt kernels - they are easy to build.
[19:56:37] <skunkworks> so it wouldn't be hard for someone having issues to try a few different ones
[19:58:59] <andypugh> pcw_home: He just admitted that he was confused by an Afrikaans translation. Next question, is there a firmware that brings out all 6 phases of the 3pwm?
[20:04:15] <jepler> andypugh: looking at PIN.vhd files in hostmot2.git, maybe PIN_TPEN4_5_48.vhd and PIN_TPEN6_6_72.vhd do? they have e.g., IOPortTag & x"00" & TPPWMTag & NTPPWMAOutPin
[20:06:41] <andypugh> _48 is typically 7i43, and _72 is 3-port PCI. He’s looking at 7i80DB, I probably need to steer him another way.
[20:09:19] <jepler> with a custom pinout you sure could fit one "six-phase" TPPWM with enable and fault (8 pins), one encoder (3 pins), and switches (3 pins) all on one DB25 header
[20:09:44] <jepler> I'm looking at hostmot2-firmware.git so there may be some examples that pcw is shipping that we're just not up to date one
[20:10:02] <jepler> we're, what, 3 years behind probably?
[20:10:27] <andypugh> I guess so. I have never been clear who “owns” that repository
[20:11:29] <andypugh> I thought there was a plan to consolidate?
[20:11:51] <andypugh> But PCW has commercial pressures.
[20:15:16] <jepler> the plan was someone on the linuxcnc side, maybe me, was going to work on getting a new source drop from peter working. then on our side we dropped the ball.
[20:15:30] <jepler> it's sure not that Peter doesn't publish the sources, they're all over his website
[20:16:07] <pcw_home> I do have a new Encoder module that has 1X mode and One Wire Threading mode
[20:17:03] <andypugh> missing tooth?
[20:17:12] <pcw_home> ( OWT=missing tooth index detection)
[20:18:05] <andypugh> Automotive target wheels have large centre bores, would be easy fit to a lathe. Use missing-tooth.
[20:18:41] <andypugh> pcw_home: I have some magnetic target wheels in my drawer at work, do you want one?
[20:19:32] <pcw_home> its a compromise for good velocity estimate (so it only counts every other tooth)
[20:21:33] <pcw_home> Yeah I wouldn't mind one (I tested with stepgen in table mode programmed to emulate a missing tooth pattern)
[20:25:20] <pcw_home> but a real signal with realistic timing errors would be a better test
[20:31:06] <jepler> zzzz so tired (adjusting to the 11 1/2 hour time difference from India to CST)
[20:33:24] <pcw_home> The way OWT works is fairly obvious: each inter-pulse time is measured and if this time becomes
[20:33:26] <pcw_home> > 1.5 x the previous inter-pulse time, a index is flagged on the next edge ( the missing tooth is one of the uncounted teeth)
[20:33:27] <pcw_home> This could of course also be added to the encoder comp for parallel port use
[20:35:19] <jepler> seems tricky when slowing to a stop
[20:36:55] <pcw_home> Not suited for tapping of course but fine for threading
[20:39:48] <jepler> reversing as a special case of stopping, yeah
[20:41:50] <pcw_home> You dont need index on reversals luckily
[20:41:52] <pcw_home> (but this is intended for one wire only for simple lathe threading with a single sensor)
[20:46:27] <andypugh> It’s something I keep meaning to add to the parport encoder.
[20:47:10] <pcw_home> Yeah it would be a good addition
[20:47:12] <pcw_home> do you see why i skip every other tooth?
[20:48:05] <andypugh> Because you can then shorten or narrow one tooth on the bull-gear of your lathe and have instant encoder + index.
[20:48:40] <pcw_home> because I dont want to make up velocity info where the tooth is missing
[20:48:58] <andypugh> I was assuming it was something clever
[20:49:11] <pcw_home> more lazy
[20:51:46] <pcw_home> (so the missing tooth is already ignored in the count/ inter-pulse time velocity data)
[20:58:52] <andypugh> Anyway, pcw_home, which FPGA cards do you build 6-output 3pwmngen firmwares for?
[21:00:18] <pcw_home> They are not common so I would probably have to make one
[21:00:19] <pcw_home> The bidirectional nature of the FANUC PWM pins makes this a bit awkward as well
[21:05:14] <andypugh> Have you seen: https://forum.linuxcnc.org/38-general-linuxcnc-questions/32273-mesa-fanuc-pwm?start=10#87324
[22:16:46] <skunkworks> the 25a amc drive makes the little emco spindle motor scream ;)
[22:17:25] <skunkworks> have touch the brushes - this motor has run clockwise forever - the reverse has a bit of brush noise