#linuxcnc-devel Logs

Jul 07 2018

#linuxcnc-devel Calendar

07:19 AM rene-dev_: KimK yes, I have everything.
07:32 AM skunkworks: rene-dev_: rene-dev_ pictures?
07:35 AM rene-dev_: skunkworksof the edm stuff?
07:38 AM rene-dev_: https://imgur.com/a/MIUZI6g
07:38 AM rene-dev_: https://imgur.com/a/Kva3mz1
07:38 AM rene-dev_: https://imgur.com/a/TK6SK6v
07:38 AM rene-dev_: https://imgur.com/a/MriFqiU
07:38 AM rene-dev_: https://imgur.com/a/ha0C5nI
08:15 AM rene-dev_: pcw_home did you read my suggestion about the pll the other day?
08:17 AM pcw_home: No, didn't see it
08:28 AM rene-dev_: pcw_home the ethercat guys added a function to adjust the servo period
08:29 AM rene-dev_: and they sync it to the slave
08:29 AM pcw_home: Yuck
08:29 AM rene-dev_: that could also be done on the fpgas
08:29 AM rene-dev_: I think its a really nice solution... your pll does the same thing, just the other way round
08:29 AM pcw_home: not good for heterogeneous systems
08:29 AM rene-dev_: yeah, agree
08:30 AM rene-dev_: we do that with the 2 chips in the stmbl, there is a pll that syncs the hv board to the clock of the low voltage board
08:30 AM pcw_home: I think linuxCNC should be in control of timing (even TOD sync is possible now)
08:31 AM rene-dev_: https://github.com/sittner/linuxcnc-ethercat/blob/master/patches/add-task-pll-functions.patch
08:31 AM pcw_home: Yeah, would not want to go that way
08:32 AM rene-dev_: I dont know why they do it, but for some systems its probably easier, or the only way to do it
08:33 AM rene-dev_: its linited to +-1% of cycle time
08:33 AM rene-dev_: https://github.com/sittner/linuxcnc-ethercat/blob/master/patches/add-task-pll-functions.patch#L68
08:33 AM pcw_home: For Ethercat (that has very tiny delays between slaves) its easy to do
08:35 AM pcw_home: as long as you want to have an external timing sourceCurrently LinuxCNC is synced to absolute time and I like that
08:36 AM rene-dev_: yeah
08:36 AM rene-dev_: whats your experience when running multiple hm2 eth boards?
08:37 AM pcw_home: I dont think I've messed much with the DPLL on multiple cards
08:38 AM rene-dev_: ok, not with dpll, just running them
08:38 AM rene-dev_: does that work reliable? normal switch?
08:38 AM pcw_home: A ran 4 cards for a month or so without issue
08:38 AM rene-dev_: ok
08:39 AM pcw_home: you need a very good host (though I can now get 4 KHz from my crappy USDT 8300 now that I turned off AMT)
08:39 AM pcw_home: I just used a normal gig switch
08:40 AM pcw_home: you need to re-arrange the hal file to use the read-request function
08:41 AM pcw_home: so you do read-request 0,1,2,3, read 0,1,2,3 write 0,1,2,3
08:44 AM pcw_home: so if you can run 4KHz on one card its likely you can to 1 KHZ with 4
08:45 AM pcw_home: (the switch acts as a concentrator to some extent)
08:46 AM rene-dev_: yeah, in theory you cant get packet loss with up to 10 slaves
09:05 AM pcw_home: I think any decent switch has full packet RX and TX buffers for every port (or more)
09:13 AM Roguish___ is now known as Roguish
09:26 AM rene-dev_: its not a lot of data anyway