#linuxcnc-devel Logs

Oct 25 2019

10:09 AM skunkworks: uh... After updating the firmware on the pi - re-installing raspian/linuxcnc using jt's directions - it seems to be running perfectly
10:12 AM skunkworks: I had 2 images - jt's install and andys image. Both exibited the following error
10:12 AM skunkworks: So - maybe it was the firmware in the pi? Time will tell
10:16 AM skunkworks: pcw_home:
10:16 AM skunkworks: http://electronicsam.com/images/greenmachine/2019-10-25-100604_1280x720_scrot.png
11:52 AM pcw_home: no big baseline shift there...
04:07 PM drdoc: I have a waaay left-field question
04:08 PM drdoc: has anybody ever attempted a HAL interface for a SCSI adapter?
04:09 PM Tom_L: i think someone has but not around much these days
04:09 PM Tom_L: iirc it was jepler but i might be mistaken
04:09 PM andypugh: I have considered a GPIB hal component…
04:09 PM drdoc: Apple used the SCSI port to interface an ethernet dongle, and I've seen systems networked on bare-metal SCSI
04:10 PM Tom_L: 7i90 iirc
04:10 PM andypugh: SCSI or IDE are both candidates for real-time machine control
04:10 PM Tom_L: wait.. that was sci
04:10 PM Tom_L: SPI
04:11 PM drdoc: andypugh: SCSI-I will do a much longer run than IDE
04:12 PM drdoc: the biggest downside is that PCIe SCSI cards are astronomically expensiv
04:12 PM drdoc: e
04:13 PM Tom_L: there already are pcie interface mesa cards
04:13 PM drdoc: yeah
04:17 PM andypugh: Now that we have Ethernet Mesa cards, a Mac port of LinuxCNC might work.
04:18 PM Tom_L: you volunteering?
04:18 PM andypugh: No. I have barely enough spare time to eat in as it is.
04:18 PM Tom_L: :)
04:20 PM Tom_L: that might make it more universal
04:25 PM andypugh: It’s an attractive idea, as OSX runs Fusion natively.
04:42 PM skunkworks: full screen youtube video - following error .0001"
04:53 PM skunkworks: interesting cause and effect... ;) How do you test your realtime control? Full screen youtube videos of course...
05:33 PM skunkworks: ugh - no. Still an issue
05:33 PM skunkworks: and it is random
05:34 PM skunkworks: andypugh: do you know of a way to 'see' cpu frequency in hal?
05:35 PM andypugh: I am not sure that HAL even knows.
05:37 PM andypugh: Some things are in clock ticks, and fperiod passed to .comps is in ns, so you could maybe calculate it. Perhaps if you were worried it was varying?
05:37 PM skunkworks: yes - something in the rpi4 is not right..
05:38 PM skunkworks: trying to figure out if the clock is changing...
05:38 PM skunkworks: I plan on outputing a square wave out of mesa gpio and scoping it (see if it varies..)
05:38 PM skunkworks: (toggled by hal)
05:40 PM skunkworks: something happens and the dpll starts to dive
05:40 PM skunkworks: http://electronicsam.com/images/greenmachine/2019-10-25-171014_1920x1080_scrot.png
05:40 PM skunkworks: no realtime missed targets - servo period taking way under 2ms...
05:42 PM skunkworks: Timings look good - but obviously something outside of hal seems to be changing...
05:46 PM andypugh: You need to talk to PCW, I suspect. phase-error might be expected to change, perhaps that’s the very job of the dpll?
05:47 PM skunkworks: we have been talking.. I think he also thinks the clock is changing or something...
05:48 PM pcw_mesa: Yeah, to me it looks like the servo thread frequency shifted beyond the DPLL lock-in range
05:55 PM pcw_mesa: so if you have a frequency counter with a long gate time you should be able to read it ( to 1 part in 2500 if you toggle bit at 2 ms intervals and have a 10 second gate)
05:58 PM pcw_mesa: you could also setup a open loop rate generator (stepgen) and measure its frequency with linuxcnc (it should be good to about 50 PPM at high (say 1 Mhz) rates)
06:18 PM skunkworks: pcw_home: open look stepgen?
06:18 PM skunkworks: remember I can only do about a 500hz base thread
06:19 PM skunkworks: are you talking the stepgen in mesa?
06:28 PM skunkworks: and then it will run for an hour without a hicough..
06:30 PM skunkworks: well - that is better - I used to be able to trigger the thing by just opening chromium... now not so puch
06:30 PM skunkworks: much
06:31 PM pcw_mesa: Yes just use a stepgen as a clock (set for 1 MHz) sample position wait say 5000 servo periods (10 seconds) and sample again
06:57 PM pcw_mesa: (so delta position after 10 seconds should be 10e6 within a 100 PPM or so)
07:29 PM skunkworks: ah - will give it a try :)
07:39 PM mozmck: andypugh: in every project I've seen that supports Mac - that is the system that causes the most problems.