#linuxcnc-devel Logs

Dec 29 2019

#linuxcnc-devel Calendar

11:48 AM skunkworks: trying andys image - because it is rip - so halcompile is there..
11:51 AM skunkworks: just running it after getting it setup - so far so good. folloing error is staying under .00004"
11:51 AM skunkworks: just going to let it run for a bit
11:51 AM jthornton: nice
11:56 AM skunkworks: Time will tell. don't know wht the difference is yet
11:57 AM skunkworks: force_turbo=1 in config.txt and your kernel line additions.
12:48 PM Tom_itx is now known as Tom_L
01:27 PM skunkworks: wow - running for a couple hours now.. Servo thread tmax is .596ms
01:47 PM JT-Shop: pcw_mesa, you around?
01:50 PM burklefoo: Yeah under an assumed name...
01:58 PM JT-Shop: incognito
02:09 PM burklefoo: somehow I've got my HexChat alternate name I never expected to get used...
02:34 PM jthornton: weird
02:34 PM jthornton: do I need a RMA for the 7i96?
02:58 PM skunkworks: What did you do!
03:01 PM jthornton: nothing the 5v socket it crooked
03:02 PM Tom_L: little heat will fix that..
03:06 PM skunkworks: well - rebooting it now - lots of chunking and high following errors
03:06 PM skunkworks: no realtime errors
03:07 PM skunkworks: same servo tmax
03:07 PM burklefoo: I was going to say just send it back with a note but I will get in trouble with the boss, so use RMA #21396
03:08 PM jthornton: lol I hear that
03:08 PM jthornton: ok
03:12 PM burklefoo: skunkworks: so you get the same following errors with Andys RPI4 image?
03:13 PM skunkworks: yes - curretnlyt a lot worse
03:19 PM burklefoo: Does the system have network connectivity
03:25 PM skunkworks: it does over wifi
03:28 PM burklefoo: I remember the default DPLL filter time was shortened to deal with systems with bad/intermittent network connections
03:30 PM burklefoo: (so it could track the rather abrupt time base adjustments that happened when NPT? adjusted the TOD)
03:30 PM burklefoo: NTP I mean
03:36 PM skunkworks: ok!
03:36 PM skunkworks: explain this
03:41 PM skunkworks: burklefoo: http://electronicsam.com/images/greenmachine/2019-12-29-213103_1920x1080_scrot.png
03:41 PM skunkworks: I have a free-running stepgen that goes 2 inches every servo thread
03:42 PM skunkworks: this was homing
03:43 PM burklefoo: The phase error indicates that the DPLL has lost sync, likely because the servo thread frequency is outside of its lock-in range
03:43 PM burklefoo: (more than a few % different than expected)
03:44 PM skunkworks: oh - so the reading is all off because of that - and why the postion difference is off
03:44 PM burklefoo: yes
03:46 PM skunkworks: the servo thread seems pretty consisstant though... still only .7ms
03:46 PM skunkworks: max
03:48 PM skunkworks: currently the timer is set to -200
03:48 PM skunkworks: is there someting I could try?
03:49 PM burklefoo: the timer just sets the sample time, if the DPLL can't lock, it suggests the servo thread period is wrong
03:50 PM skunkworks: so my way of checking the servo thread won't work because by the time the shit hits the fan - the pll acts up
03:51 PM burklefoo: you could disable the DPLL
03:51 PM skunkworks: so really - need a hal way of measuring the actual servo thread time (not what linuxcnc says it is)
03:51 PM skunkworks: I think I did - but I could try it again
03:51 PM skunkworks: how do you disable it? set the timer to 0?
03:51 PM burklefoo: you need to set the stepgen timer number to -1
03:52 PM skunkworks: ok
03:52 PM burklefoo: (thats also the default if you dont set it)
03:52 PM skunkworks: ok
03:53 PM burklefoo: you coul als try settin the DPLL time constant shorter (say 200)
03:53 PM burklefoo: could also
03:55 PM skunkworks: packet errors seem to be true all the time
03:55 PM skunkworks: does that seem right?
03:57 PM skunkworks: that can't be right
04:00 PM skunkworks: burklefoo: http://electronicsam.com/images/greenmachine/2019-12-29-215213_1920x1080_scrot.png
04:11 PM burklefoo: no, that's just weird...
04:13 PM burklefoo: do you have the recv-packet timeout set shorter than the default?
04:16 PM skunkworks: packet read timeout?
04:16 PM skunkworks: 60
04:19 PM skunkworks: changed network cable..
04:19 PM skunkworks: without dpll I get instant.0026" following error
04:19 PM skunkworks: sounds rought
04:19 PM skunkworks: rought
04:19 PM skunkworks: rough
04:20 PM burklefoo: An this is 2 ms?
04:21 PM skunkworks: yes
04:21 PM burklefoo: well that should not timeout until 1.2 ms
04:22 PM skunkworks: with the dpll working - I get less than .00005"
04:22 PM skunkworks: let me turn dpll back on
04:22 PM burklefoo: it may be that the way hm2_eth calculates the timeout fails on the RPI4 for some reason
04:25 PM burklefoo: does setting the timeout to 90 reduce the packet errors?
04:25 PM burklefoo: or setting it to say 1200000
04:26 PM skunkworks: I am not getting any packet errors since I switched network cables
04:26 PM skunkworks: odd
04:28 PM skunkworks: it ran for hours before the reboot with no issues
04:28 PM skunkworks: then crap - and I noticed the packet errors
04:28 PM skunkworks: swapped out net work cables and golden
04:28 PM skunkworks: so far
04:29 PM skunkworks: trying a reboot
04:34 PM burklefoo: weird
05:15 PM skunkworks: well - have to play with it more later
06:33 PM skunkworks: someone on youtube is using a 7i92 and pi and getting following errors too - but I don't know the details.
07:22 PM skunkworks: there is also still an iptables error when you first start linuxcnc after a fresh reboot
07:31 PM skunkworks: I need to check to make sure the iptable stuff is incerted
07:35 PM skunkworks: Have to bring the fluke network tester home and test the cable