#linuxcnc-devel | Logs for 2016-09-01

[07:17:57] <lair82> Good Morning Guys, any idea as to what this fault may be telling me? bad 7i73? bad cable? https://app.box.com/s/watg5likmdz2srb3k0kjpp5rxn4hnwr2
[08:13:57] <skunkworks> zlog
[08:27:27] <lair82> pcw_home, you around yet this morning?
[09:20:48] <pcw_home> Yeah
[09:22:57] <lair82> Hey pcw_home, any idea as to what this fault may be telling me? bad 7i73? bad cable? https://app.box.com/s/watg5likmdz2srb3k0kjpp5rxn4hnwr2
[09:23:13] <pcw_home> Lair62: Those errors seem like a lost connection
[09:23:15] <pcw_home> I would check power at both ends (5V at 7I44? and 7I73 end)
[09:24:17] <seb_kuzminsky> the beaglebone kernel guy, robert nelson, is shipping vanilla and rt-preempt kernel debs now, that's nice
[09:24:27] <seb_kuzminsky> Get:1 http://repos.rcn-ee.com/debian/ jessie/main linux-image-4.4.19-ti-rt-r41 armhf 1jessie [22.4 MB]
[09:25:25] <lair82> OK, I will watch it, they are running the machine now. I swapped out the 7i73, thinking that may be the problem, but wasn't sure.
[09:27:51] <skunkworks> seb_kuzminsky, neat!
[09:28:08] <skunkworks> so linuxcnc prime should run on it?
[09:28:40] <pcw_home> lair82: It doesn't look like noise, I suspect you would get the same errors if you just disconnected the serial cable when running
[09:30:25] <seb_kuzminsky> skunkworks: the jessie uspace debs should run i think, yeah
[09:30:38] <seb_kuzminsky> i'll try that next
[09:33:08] <skunkworks> seb_kuzminsky, you have a BBB?
[09:48:05] <seb_kuzminsky> skunkworks: yeah i recently dug it out of a drawer
[09:48:30] <seb_kuzminsky> it used to be the original buildbot armhf buildslave, until i switched to the u3
[09:49:23] <seb_kuzminsky> i'm planning to write hostmot2 firmware for the bbb pru
[09:49:25] <skunkworks> have the u3's been staying cool then?
[09:49:52] <skunkworks> I just don't like how crappy the video is on the BBB
[09:50:09] <skunkworks> (tool path and such)
[09:51:10] <skunkworks> That sounds cool!
[09:56:19] <jepler> oops, a spacex rocket exploded during a test fire. no injuries, but apparently the payload (couple of satellites) was mounted during the test fire and assumed lost
[09:56:40] <cradek> heck
[09:59:05] <jepler> smoke plume visible on doppler radar, next to the last video https://techcrunch.com/2016/09/01/a-spacex-falcon-9-rocket-just-exploded-at-cape-canaveral/
[10:00:18] <cradek> maybe they just decided to help the world by blowing up facebook's satellite
[10:00:26] <Tom_itx> not during launch?
[10:01:03] <Tom_itx> some maintenance guy smoking under it??
[10:03:11] <jepler> this wasn't the re-used first stage, was it?
[10:12:13] <seb_kuzminsky> jepler: that one's not until NET november i think
[10:14:35] <seb_kuzminsky> skunkworks: yeah, it's got slow graphics, it's low on RAM, not my favorite computer
[10:14:45] <seb_kuzminsky> but it's $50 and it's got the PRUs...
[10:14:50] <skunkworks> right
[10:15:46] <seb_kuzminsky> haha i ran latency-test and it said "unexpected latency, run latency-test and resolve before continuing"
[10:17:08] <skunkworks> circular reference!
[10:17:18] <skunkworks> nobase
[10:18:20] <seb_kuzminsky> jepler: isn't your little router a zeonbot? is it from this company? http://www.zenbotcnc.com/cnc-routers
[10:19:05] <jepler> seb_kuzminsky: that may be the distant descendant. they were just selling on ebay when I bought mine.
[10:19:36] <cradek> I forgot about zenbot!
[10:20:08] <cradek> clearly you've decided ordering pcbs is easier than cutting them
[10:29:41] <seb_kuzminsky> a friend wants to buy a woodworking router for 8x4 sheets of plywood and that was the least expensive one he could find
[10:30:01] <seb_kuzminsky> it's a shopbot-sized machine with a gantry made of hdpe...
[10:30:17] <cradek> did he include paying for your time for rebuilding it so it's usable?
[10:30:29] <seb_kuzminsky> well i can apt-get install an rt-preempt linux-image and linuxcnc-uspace on my bbb, and tklinuxcnc starts, so that's something
[10:30:46] <seb_kuzminsky> i knew procrastinating for a couple of years would make that board easier to deal with!
[10:31:17] <seb_kuzminsky> cradek: i told him to just come use the shopbot at the hack space that we already fixed up, but he's set on having one in his garage
[10:31:51] <cradek> I understand that urge, but ouch
[10:40:22] <seb_kuzminsky> does mesa make stepper drivers, or is that still gecko territory?
[10:43:58] <jepler> seb_kuzminsky: mesa used to have one or two stepper drivers, but I don't think they're offered anymore.
[10:44:27] <jepler> ah there's this thing http://store.mesanet.com/index.php?route=product/product&product_id=143
[10:44:44] <seb_kuzminsky> ooh,
[10:44:50] <seb_kuzminsky> "look on their website", novel idea
[10:45:21] <jepler> manual link seems to be missing? http://www.mesanet.com/pdf/motion/7i32man.pdf
[10:45:34] <pcw_home> we have some little ones that work with SoftDMC (closed loop stepper)
[10:45:48] <jepler> anyway it doesn't seem that compelling a device, you only get 2 steppers on 1 50-pin connector
[10:45:49] <seb_kuzminsky> that's ~= to 1/2 of a G540, at ~ 1/3 of the price, nice
[10:46:03] <jepler> and I think you need a FPGA firmware that does the PWM for you (?)
[10:46:27] <cradek> buy a g540 so you can test my PR
[10:46:31] <skunkworks> too bad pcw_home isn't here
[10:46:38] <skunkworks> ;)
[10:46:41] <jepler> quadrature feedback? interesting.
[10:47:36] <skunkworks> I have not found hardware yet that exibits the 'os turning off EPP mode'
[10:47:47] <skunkworks> It could have been a ubuntu thing...
[10:48:03] <pcw_home> Yes SoftDMC supports servo mode closed loop steppers
[10:48:26] <skunkworks> Currently I just found a bunch of cards that don't seem to do hardware epp
[10:49:40] <pcw_home> I thought any parallel card <20 years or so old did EPP
[10:50:14] <skunkworks> it is funny that mach says they don't want to support mesa hardware because they would have to write the driver / charge for it. but they are doing exactly that for galil.
[10:52:00] <skunkworks> pcw_home, all I have here are moschip boards
[10:52:35] <skunkworks> I cannot seem to force them into EPP where they should exibit totem pole output
[10:52:59] <pcw_home> Those do support EPP (as far as signal drive issues go) some just have a protocol bug
[10:54:03] <skunkworks> well then I don't know how to force them into that mode. The normal way - like how the hostmot driver does - doesn't seem to work.
[10:54:26] <skunkworks> (it was added to the printer port driver by cradek
[10:54:51] <skunkworks> whatever I do it seems to have just a oc setup
[10:56:00] <skunkworks> I can see the normal printer port switch for epp to legacy. (totem pole to oc with pullup)
[10:56:14] <skunkworks> if I am making any sense
[10:56:46] <skunkworks> (these are all pci cards)
[11:01:42] <pcw_home> what chip do they use? I think I have the data sheet for some
[11:05:39] <skunkworks> mcs9805cv
[11:07:05] <jepler> https://4donline.ihs.com/images/VipMasterIC/IC/ASIX/ASIXS00037/ASIXS00037-1.pdf
[11:11:59] <jepler> pin nSTROBE is noted as OD_O (PU) [open drain output, pull-up) on page 10
[11:13:52] <skunkworks> pin 16 is the charge pump on the
[11:14:40] <skunkworks> wait 17?
[11:15:28] <jepler> nINIT and nSLCTIN (pins 16 and 17 on db25 parallel port) are also both OD_O (PU)
[11:16:07] <jepler> maybe the world needs an itty bitty card with a couple of DB25 connectors on it, a USB plug, and a totem pole driver for some of these problematic signals
[11:16:08] <skunkworks> so did gekco pick a pin that doesn't do totem pole on a lot of hardware?
[11:17:06] <skunkworks> 16
[11:17:53] <skunkworks> I only know of one that exibits this issue (g540)
[11:19:17] <pcw_home> Actually almost all interfaces that use OPTOs (and no parallel port side power) violate parallel port specs
[11:19:56] <pcw_home> really only TTL levels are legal (2.4V high .4V low)
[11:22:31] <pcw_home> a better data sheet with register descriptions:
[11:22:33] <pcw_home> http://freeby.mesanet.com/9805.pdf
[11:23:26] <skunkworks> This is the onboard printer port
[11:23:27] <skunkworks> http://electronicsam.com/images/matsuura/20160721_054539.jpg
[11:23:43] <skunkworks> This is the pci card,
[11:23:44] <skunkworks> http://electronicsam.com/images/matsuura/20160721_054558.jpg
[11:24:09] <skunkworks> I can make the onboard card look like the pci card by changing it to legacy in the bios
[11:24:16] <skunkworks> or spp or whatever it is called
[11:25:01] <archivist> add a pullup?
[11:25:33] <skunkworks> I think I put a few hundred ohm to ground to mimic the g540
[11:26:10] <archivist> pull up to 5v because open collector
[11:26:35] <skunkworks> right - but that isn't what the 540 has - (although that is the fix some have done...)
[11:26:46] <skunkworks> Like kwallace2
[11:27:01] <skunkworks> joe blow user doesn't want to do something like that.
[11:27:52] <archivist> so many violate specs though
[11:34:06] <archivist> I agree with jepler about a db25 to 25 with usb for the 5v with some buffers on it pcb
[11:34:41] <cradek> I bet that would fix my pluto problems too
[11:35:11] <cradek> you could call it the "make any shitty parallel port hardware work right" board
[11:35:53] <skunkworks> or just by a mesa board...
[11:38:13] <archivist> better name BOBs you're uncle
[11:39:49] <skunkworks> heh
[11:51:23] <jepler> https://emergent.unpythonic.net/files/sandbox/byo.png
[11:51:59] <skunkworks> oh wow ;)
[11:52:24] <skunkworks> still using eagle?
[11:52:29] <jepler> yeah
[11:52:33] <cradek> doesn't it need about twice as many amplifiers?
[11:52:45] <jepler> depends on your goals
[11:52:57] <skunkworks> I think jepler is assuming that the rest are totem pole
[11:53:04] <cradek> ah
[11:53:12] <jepler> right
[11:53:16] <skunkworks> I have not checked that actually..
[11:53:20] <skunkworks> I will
[11:53:32] <jepler> this puts drivers on pins 1, 14, 16, 17
[11:55:23] <jepler> data bus (2..9) and inputs (10-13, 15) are just passed through
[11:55:24] <skunkworks> for the g540 then it could be as simple as making a cable that swaps 16 with one of the data pins.. ;)
[11:56:03] <cradek> jepler: it would sure be fun to try that on the pluto
[11:56:44] <jepler> "fun" sure
[11:58:20] * archivist giggles
[11:58:22] <jepler> hmm you still use old eagle don't you
[11:58:44] <cradek> who me?
[11:58:47] <jepler> cradek: yeah you
[11:58:54] <cradek> I don't know what I use anymore
[11:58:57] <jepler> hah
[11:59:07] * archivist never ever used eagle
[11:59:07] <cradek> probably whatever installs when I apt-get install eagle
[11:59:33] <skunkworks> I got pretty darn good with eagle - but I have not used it in years now.
[11:59:41] <archivist> kicad is the new kid on the open sauce block
[12:03:04] <jepler> cradek: anyway if you are motivated I'd be happy to send you what I just slopped together, but I'm not sure it's useful to you
[12:03:36] <jepler> it's in eagle 7.1.0 format
[12:04:31] <jepler> perfect chance for you to try out the oshpark service, about $7 for 3 copies of this board, but you have to be patient on the turnaround.
[12:04:44] <seb_kuzminsky> well, on the bbb, "apt-get install linux-image-mumble-rt linuxcnc-uspace" just works, axis starts (after switching to libgl1-mesa-swx11), that's all pretty easy
[12:05:19] <jepler> seb_kuzminsky: nice!
[12:05:43] <jepler> seb_kuzminsky: when I was toying recently after jmk's interest, I used a MK image with xenomai realtime and it was also fairly smooth
[12:05:54] <jepler> but that needs master branch of course
[12:06:57] <jepler> how's latency-histogram look?
[12:09:47] <seb_kuzminsky> i'll have to try that when my X display is closer to the board
[12:10:25] <seb_kuzminsky> it doesn't work when it has to hop over wifi to my home network, then over my DSL to $DAYJOB, i can tell you that
[12:10:35] <seb_kuzminsky> i'll look tonight
[12:12:46] <jepler> sure, of course
[12:14:46] <seb_kuzminsky> the gnu-pru examples Just Work on the RN rt-preempt kernel, that's nice
[12:14:50] <seb_kuzminsky> bbl
[13:03:36] <mozmck> Hmm, I remember seeing some work on absolute encoder support - is that complete? https://forum.linuxcnc.org/forum/10-advanced-configuration/31496-getting-soft-limits-to-work-without-homing
[13:08:25] <pcw_home> Master (JA) only I think
[14:04:55] -linuxcnc-github:#linuxcnc-devel- [13linuxcnc] 15nicokid commented on issue #86: Ok @jepler I'm controlling also the issue #57.... 02https://github.com/LinuxCNC/linuxcnc/issues/86#issuecomment-244171362
[14:46:51] <KGB-linuxcnc> 03Norbert Schechner 05master 43a836c 06linuxcnc 10src/emc/usr_intf/gmoccapy/getiniinfo.py gmoccapy_2_0_XX - get_ini_info - missing entries leaded to a python exception * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=43a836c
[14:55:53] <KGB-linuxcnc> 03Norbert Schechner 05master 6244dc3 06linuxcnc 10src/emc/usr_intf/gmoccapy/gmoccapy.py 10src/emc/usr_intf/gmoccapy/release_notes.txt gmoccapy_2_1_2 - spindle label has not been updted * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=6244dc3
[16:07:17] <seb_kuzminsky> rt-preempt latency on the bbb is unusable: http://highlab.com/~seb/linuxcnc/bbb-latency.png
[16:07:45] <seb_kuzminsky> that's with X forwarding to a machine on the same network over ssh
[16:08:11] <cradek> that's unfortunate
[16:08:39] <cradek> umm that's .05 seconds?
[16:09:51] <cradek> that might be the worst result we've ever seen
[16:16:31] <jepler> seb_kuzminsky: ow
[16:20:00] <jepler> "Using POSIX realtime" on the terminal?
[16:21:28] <seb_kuzminsky> jepler: yeah, and rtapi_app is setuid root
[16:27:05] <seb_kuzminsky> and i ran it with --nobase
[16:27:18] <seb_kuzminsky> and i dont have enough swap space on that SD card to build linuxcnc :-/
[16:39:42] <seb_kuzminsky> this kinda makes me want a textmode latency-test
[16:40:35] <jepler> cyclictest from package rt-tests?
[16:40:44] <seb_kuzminsky> i'll try it
[16:46:03] <seb_kuzminsky> well that looks better:
[16:46:05] <seb_kuzminsky> T: 0 ( 6560) P: 0 I:1000 C: 77520 Min: 41 Act: 472 Avg: 135 Max: 26328
[16:48:33] <jepler> that's in us, not ns
[16:48:56] <jepler> (by default at any rate)
[16:51:08] <seb_kuzminsky> oh bummer
[16:51:55] <jepler> -N shows values in ns
[17:04:20] <seb_kuzminsky> some dude claims they got worst-case scheduling latency of 55 us on the bbb with 3.14-rt
[17:11:08] <jepler> that would be nice
[17:30:35] <seb_kuzminsky> cyclic.4.4.19-ti-rt-r41:# Min Latencies: 00020
[17:30:35] <seb_kuzminsky> cyclic.4.4.19-ti-rt-r41:# Avg Latencies: 00055
[17:30:35] <seb_kuzminsky> cyclic.4.4.19-ti-rt-r41:# Max Latencies: 00183
[17:46:19] <jepler> even 183 << 26328
[19:11:58] <seb_kuzminsky> yeah, 183 is usable, +inf (or whatever) is not
[19:35:33] <jepler> try xenomai, it works with uspace in master branch