#linuxcnc-devel | Logs for 2014-11-11

[09:37:35] <CaptHindsight> http://psha.org.ru/ or you end up at the Pennsylvania Speech-Language-Hearing Association website
[09:38:06] <skunkworks> heh
[09:38:10] <skunkworks> thanks
[10:05:20] <skunkworks> pcw_home, what where you looking at with latency>
[10:06:02] <pcw_home> The J1800-D2H on RTAI
[10:07:08] <pcw_home> or are you asking about my general distrust of the latency test...
[10:09:00] <archivist> I agree with the distrust
[10:09:43] <archivist> I did spend some time reviewing it and think it needs some fixing
[10:15:58] <pcw_home> I think it gives quite optimistic timing on the servo thread mainly (its shows say 8-10 us on a D525, but a D525 has up to 75 random usec delays in the
[10:15:59] <pcw_home> worst linuxcnc functions (maybe ones the are delayed by hardware resources being blocked or memory/caching issues)
[10:18:07] <pcw_home> you can get an idea of the magnitude of this by plotting function times (and looking at tmax'es)
[10:18:09] <pcw_home> (some tmaxes need to be reset before you start since they seem to have large startup delays_
[10:33:46] <KGB-linuxcnc> 03Sebastian Kuzminsky 05liblinuxcnc-ui c92f0d2 06linuxcnc 10tests/halui/jogging/test-ui.py 10tests/halui/mdi/test-ui.py halui tests: better logging * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=c92f0d2
[10:53:21] <skunkworks> pcw_home, not moving the mouse.. http://electronicsam.com/images/KandT/testing/Screenshot%20-%2011112014%20-%2006:32:21%20PM.png
[10:53:38] <skunkworks> moving the mouse. http://electronicsam.com/images/KandT/testing/Screenshot%20-%2011112014%20-%2006:32:56%20PM.png
[10:54:46] <pcw_home> Yeah, the 10 Hz Axis update rate is pretty cleary visible :-)
[10:56:03] <pcw_home> same thing happens with RTAI...
[10:56:46] <skunkworks> so 325k at 2ghz.. what is that?
[10:57:42] <pcw_home> 160 usec?
[10:58:18] <skunkworks> that seems really good
[10:59:39] <skunkworks> from what I have seen...
[10:59:48] <pcw_home> eventually it will get higher (flash videos seem to be the worst)
[11:00:37] <skunkworks> I can reset tmax and see what it creeps up to.
[11:00:48] <skunkworks> running a gcode loop
[11:01:15] <pcw_home> the write tmax seems to get set high when starting
[11:02:17] <pcw_home> but try on RTAI and mesa PCI or parallel port, you will see the same thing (shorter times but same spikes)
[11:03:23] <pcw_home> best Ive seen so far for Ethernet is a fast CPU, but J1800,1900 are fine at 1 KHz
[11:04:30] <pcw_home> +DPLL to take the sampling jitter out of the position loop
[11:09:18] <pcw_home> did you try the DPLL yet?
[11:09:59] <skunkworks> without reseting it - it is at 399k
[11:10:01] <skunkworks> no
[11:10:40] <pcw_home> it reduces the stepgen FE by a factor of at least 10
[11:10:43] <skunkworks> wrie is at 35k
[11:11:20] <skunkworks> I am running the J1900 right now at 2khz
[11:53:11] <mozmck> what brand of board is that skunkworks?
[11:53:42] <skunkworks> gigabyte J1900n-D3v
[11:53:59] <skunkworks> it is a quad core with 1 pc1 and 2 nics
[11:54:32] <mozmck> ok, thanks! Is that one that you had to update the BIOS to get linux to boot?
[11:54:44] <skunkworks> I don't remember...
[11:55:00] <mozmck> ok
[11:55:05] <pcw_home> I did on mine (to V3)
[11:55:25] <pcw_home> (but mine is a Gigabyte D2H)
[11:55:26] <mozmck> ah, ok. Is there a different brand that is better?
[11:55:58] <pcw_home> My ASrock J1900 booted linux without issue
[12:00:10] <pcw_home> this one looks pretty good if you want more than one PCI slot:
[12:00:12] <pcw_home> http://www.newegg.com/Product/Product.aspx?Item=N82E16813157565&cm_re=j1900-_-13-157-565-_-Product
[12:00:32] <CaptHindsight> heh Flash videos, the real time killer
[12:01:07] <pcw_home> yeah, they are the worst thing I have found so far
[12:18:32] <kwallace3> Good morning. I'm looking at some servos on eBay, such as: 111484598705 . My problem is that a common option, R2 is a resolver primary encoder instead of a hall encoder. See page label 25 and 32 here: http://www.kollmorgen.com/en-us/products/motors/servo/serviced-motors/r-series-pac-sci-/_literature/servo_motor_selection_guide_en-en_2003.pdf/ which says R2 = "frameless resolver". Can the Mesa or Pico resolver adapters be used with this kind of
[12:24:10] <kwallace3> ... It looks like frameless has to do with a mechanical configuration rather than electrical?
[12:35:33] <CaptHindsight> yes, no mechanical frame
[12:36:21] <kwallace3> http://www.kollmorgen.com/en-us/products/motors/direct_drive/serviced_motors/_application_notes/frameless_brushless_motor_installation_en-us.pdf/
[12:43:51] <pcw_home> heres a bunch of pretty big servos:
[12:43:53] <pcw_home> http://www.ebay.com/itm/Lot-7x-GE-Fanuc-AC-Servo-Motor-w-Heidenhain-Interpolation-Digitizing-Fan-/201211973364?pt=LH_DefaultDomain_0&hash=item2ed92b0ef4
[12:55:07] <kwallace3> I have a Pac Sci R46 I got a few years back on eBay just because it was cheap. I noticed that it has the same NEMA 42 mount as my Shizuoka steppers, so I'm thinking I might be able to find two more servos to replace the steppers. They have to be NEMA 42 and direct drive because the motor passes partially into a cutout in the table base at the -X limit.
[12:55:51] <kwallace3> oops, +X limit.
[12:58:57] <kwallace3> The R46 I have has the Hall encoder, so should be easy to drive. Then I need to add a quadrature encoder for position in the same 4.2 inch square.
[13:01:08] <pcw_home> you probably want 8 pole or more if you are direct driving the ballscrews
[13:01:41] <pcw_home> dont like those 7HP Fanucs?
[13:02:39] <pcw_home> (those are 41A cont so a about a 175A drive needed...)
[13:08:52] <kwallace3> The Fanucs are about 5 HP too big. My other option, is to hang the motor out another six inches. Then I can add pulleys for the motor and encoder. That may be a better way to go, but I'm bumping into current motor as it is. It's a good thing I'm not planning to have kids.
[13:11:08] <pcw_home> :-)
[13:11:09] <pcw_home> for direct drive the trick is finding low RPM (and high torque) motors
[13:11:40] <pcw_home> the fanuc 8 pole motor are often 2000 RPM which is not too bad
[14:08:23] <andreas> hi
[14:09:20] <andreas> I have a question about compiling machinekit, im here on the right place?
[14:10:16] <cradek> sorry no, they have a separate list
[14:12:00] <andreas> do you know, they have a irc channel?
[14:12:10] <cradek> not that I know of
[14:12:27] <andreas> oke thx
[14:27:28] <CaptHindsight> looks like they registered a channel but it's empty
[17:08:13] <seb_kuzminsky> huh, i can repro that "halui doesn't send the mdi command" problem that we saw in the liblinuxcnc-ui branch on the armhf buildslave
[17:08:23] * seb_kuzminsky puts his debugging pants on
[17:15:47] <pcw_home> Are those DS waders?
[17:23:01] <seb_kuzminsky> http://static.guim.co.uk/sys-images/Guardian/Pix/pictures/2014/8/19/1408461775551/Thames-Water-worker-remov-014.jpg
[17:25:04] <pcw_home> :-)
[17:25:36] <pcw_home> hmm stepconf gets confused by uspace
[17:28:50] <seb_kuzminsky> is it the probe_parport problem, or something else?
[17:29:20] <pcw_home> well it want to make a sim config
[17:29:25] <pcw_home> wants
[17:58:01] <cmorley> pcw_home: if it's recent build look that the 'create simulated hardware' checkbox is unchecked. Seems the default is to create a simulated config
[17:58:18] <cmorley> It's on the start page
[18:50:40] <pcw_home> OK Let me try
[18:53:45] <pcw_home> OK thats all it was though I get a error/notifier popup
[18:53:47] <pcw_home> "Linux parport #0 @0x378 0x0"
[18:55:59] <pcw_home> cmorely did you happen to see this:
[18:56:00] <pcw_home> http://www.linuxcnc.org/index.php/english/forum/39-pncconf/28539-5i257i76-spindle-not-working
[18:56:02] <pcw_home> (two typos in pncconf boilerplate)
[18:56:23] <jepler> that one's a debugging message that should be removed
[18:56:55] <jepler> it juts says that linux knows about a parport at 0x378, no extended regsters (0x0); its linux parport number i s0
[18:56:58] <jepler> is 0
[18:58:11] <KGB-linuxcnc> 03Jeff Epler 052.7 2675514 06linuxcnc 10src/rtapi/uspace_rtapi_parport.cc parport: remove debugging message * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=2675514
[18:58:17] <pcw_home> OK looked like one...
[18:58:18] <pcw_home> hmm wonder if I have parallel port cable here
[20:52:38] <jepler> http://ssvb.github.io/2014/11/11/revisiting-fullhd-x11-desktop-performance-of-the-allwinner-a10.html
[21:10:01] <skunkworks_> I bet I can make jepler cry... http://pastebin.com/YL9hpEWL
[21:13:12] <jepler> while(True):
[21:13:21] <jepler> normal Python style would be just while True:
[21:13:26] <jepler> cv2.destroyAllWindows
[21:13:34] <jepler> missing parens mean the function won't actually be called
[21:16:03] <jepler> of course it would benefit from being factored out into functions, but just what functions I'm not going to tell you
[21:16:52] <skunkworks_> heh - to I get extra points for it actually working?
[21:16:57] <skunkworks_> Do..
[21:17:52] <skunkworks_> Hmm - The cv2.destroyAlllWindows was taken right from opencv's example
[21:18:14] <jepler> that doesn't necessarily make it right
[21:19:26] <jepler> but yes, you get points for it working
[21:20:03] <jepler> roland at txrx will want this code once he gets further along on his combined PCB scan-exposing / drilling rig
[21:20:13] <jepler> I think the process involves placing the board at least 4 times
[21:20:47] <jepler> .. but the machine wasn't even moving on two axes by the time we left, let alone drilling, exposing, and imaging
[21:21:57] <skunkworks_> all this does is detect circles of a given range of sizes. I have not looked to see what other shapes opencv natively finds..
[21:22:44] <skunkworks_> (and allows you to average multiple samples)
[22:06:48] <skunkworks_> If I can actually get the camera working for more than a few seconds - I will post it on the list
[22:20:26] <seb_kuzminsky> that's cool, skunkworks_
[22:20:51] <seb_kuzminsky> hmm, 1764 iterations of the halui/mdi test and no more failures :-/
[22:21:18] <skunkworks_> yay?
[22:21:23] <seb_kuzminsky> boo
[22:21:28] <skunkworks_> oh
[22:21:32] <seb_kuzminsky> there's a bug in there, it's just hiding really well :-/
[22:22:05] <seb_kuzminsky> i know! i'll ignore it and go frivolously play with the robot arm at the hackspace & drink beer with my nerdy buds! :-)
[22:22:11] <seb_kuzminsky> bye!
[22:40:50] <cmorley> pcw_home: yes thanks I saw it - Though am not sure what version of linuxcnc he is using. In master I am working on something that will cover this.
[23:00:17] <CaptHindsight> skunkworks_: http://www.youtube.com/watch?v=FirtmYcC0Vc OpenCV Python Homography
[23:08:03] <CaptHindsight> http://www.youtube.com/watch?v=bSeFrPrqZ2A Real-Time Object Tracking