#linuxcnc Logs

Sep 22 2020

01:03 AM Deejay: moin
04:02 AM Loetmichel: maaan... my home "servers" PSU is dead again. third one to die. I think that SFF lenovo desktop that i use as server isnt meant to host half a dozen 2tb WD black :-(
05:05 AM JT-Cave: morning
05:22 AM XXCoder: hey
06:09 AM Tom_L: morning
06:10 AM XXCoder: hey
06:11 AM Tom_L: JT-Cave, i figured out the estop code
06:13 AM JT-Cave: cool
06:20 AM JT-Cave: looks like we might get some showers the next two days
06:24 AM Tom_L: 59F 73 hi
06:25 AM Tom_L: says we might get some but doesn't look so promising
06:27 AM Tom_L: JT-Cave, did you ever get updated to 2.8?
06:27 AM Tom_L: or were you having dependency issues still
06:28 AM JT-Cave: yea and no the last issue was the menu in the Mint desktop but I think someone has solved that mystery
06:28 AM Tom_L: what machine was this on?
06:30 AM JT-Cave: both debian 9 and 10
07:20 AM robotustra: morning
09:06 AM jymmmm: morning
09:47 AM * flyback listens to some ponyhead
09:47 AM flyback: ponyhead owns your "canuck"
09:47 AM * flyback dropkicks gregcnc
09:53 AM roycroft: so from february through sunday we had in total just over 50 case of covid-19 in my county
09:53 AM roycroft: we had 70 new cases yesterday
09:53 AM roycroft: almost all of them were 20 somethings who went back to partying
09:53 AM flyback: what country is that
09:54 AM roycroft: county, not country
09:54 AM flyback: ah
09:54 AM roycroft: lane county oregon
09:54 AM roycroft: school was set to start next monday
09:54 AM roycroft: now it's postponed again because we no longer qualify due to the surge of cases yesterday
09:55 AM roycroft: thanks, dumbasses!
10:00 AM flyback: yeah well people are stupid canucks
10:01 AM roycroft: i'm not aware of any foreign nationals amongst the 70
10:01 AM roycroft: but that statistic is not reported
10:01 AM roycroft: i doubt there were many, of any, though
10:30 AM Loetmichel: hehe. $me just removed a few M12 bolts from the wall of the measurement chamber. "UUUUUMPH...HUUURGH... PLINGGGGG... MAAN, who tightended them?" Boss: "That would have been Mr.Loetmichel a few years ago. :-)"
10:35 AM gregcnc: still covid eh... bye
10:35 AM pjm: btw there is a company in the USA called major tool and machine, some big ass CNC's there www.youtube.com/watch?v=ykV1SAAMDw4 20ft + beds etc, very interesting to see their machines
11:37 AM CaptHindsight: https://www.youtube.com/watch?v=vLcOeH1CTMQ DeskProto CAM for Linux, as 64 bits AppImage file
11:43 AM CaptHindsight: skunkworks: force_turbo=1 should always be enabled in the Rpi4 config.txt file
11:46 AM CaptHindsight: I have used arm_freq to over 2150 along with over_voltage=6 ( this sets the warranty bit) with a proper heatsink but it really hasn't lowered the latency scores
11:48 AM CaptHindsight: to lower the latency on the Rpi4 he installed all the software rendering modules and killed hardware accell
11:50 AM skunkworks: ah - that is what i was wondering - how he killed the hardware accel
11:50 AM CaptHindsight: looking for his blacklist file
11:50 AM Tom_L: be nice if you posted it somewhere
11:51 AM CaptHindsight: I will you spaz
11:51 AM CaptHindsight: (will be using only 80's terms today)
11:51 AM Tom_L: reached spaz status so soon...
11:52 AM Tom_L: just happened to glance over.. working on other stuff
11:55 AM CaptHindsight: blacklist-gpu.conf https://pastebin.pl/view/2f57585f
11:57 AM CaptHindsight: he has this in the /rootfs/etc/modprobe.d/
11:59 AM CaptHindsight: the max core frequency for the rpi4 set by arm_freq= seems to vary a bit by kernel and options set
12:00 PM CaptHindsight: same Rpi4 worked with 2150 with the LCNC 2.8 img but only to 2100 with some other Rpi images we had
12:02 PM CaptHindsight: skunkworks: I'll find out later what else is required
12:02 PM CaptHindsight: to kill hardware accel
12:03 PM skunkworks: CaptHindsight: thank you!
12:04 PM Tom_L: take good notes :) i'm in and out here right now
12:12 PM CaptHindsight: the Rpi4 seems fine if you want to combine it with a mesa FPGA
12:13 PM rs[m]: why should higher clock speed reduce latency
12:13 PM rs[m]: overclocking could lead to more frequent spurious throttling due to thermal issues, that does affect latency
12:13 PM CaptHindsight: for software stepping I'd help the "Russians" fix the integrated microcontroller driver for the Allwinners used in the OrangePi's
12:14 PM CaptHindsight: rs[m]: throttling should be off
12:15 PM rs[m]: CaptHindsight: you can't completely turn off throttling on the rpi, nobody knows what the main cpu on this thing is up to
12:15 PM CaptHindsight: yes, that is the problem with the Rpi4
12:16 PM rs[m]: (nobody means nobody outside NDA broadcom sphere)
12:16 PM CaptHindsight: if force_turbo=1 actually did what it says it does that would be nice
12:16 PM CaptHindsight: that is why the init should be u-boot with throttling removed from even being enabled
12:17 PM CaptHindsight: which is another reason why the Allwinners are better for software stepping, they maybe be used with u-boot
12:18 PM rs[m]: i'm not sure about the rpi4, but the 3 series have the mini uart baudrate generator coupled to the clock that ised to generate cpu core clock, so using the mini uart automatically locked the core cpu clocks
12:18 PM rs[m]: (else the baudrate would change with a clock change, that would not be easy to hide)
12:20 PM CaptHindsight: Allwinners over Broadcom for use with RT
12:20 PM CaptHindsight: but the Rpi's won the popularity contest
12:22 PM rs[m]: this https://github.com/kinsamanka/PICnc-V2 would also work with linuxcnc and could be ported to your $1 ARM µC cortex-m0 du jour
12:22 PM CaptHindsight: Allwinner also use Mali for the versions we'd be using
12:22 PM CaptHindsight: so open source for now until nVidia gets their paws on it
12:22 PM rs[m]: so if you design a breakoutboard, just add some 1$ microcontroller, and don't worry about step timings
12:26 PM CaptHindsight: rs[m]: do you know how much the board costs vs the just the copper on the PCB?
12:26 PM rs[m]: don't understand your question
12:27 PM CaptHindsight: the complete assembled breakout board vs just the copper on the PCB?
12:30 PM rs[m]: still don't understand. which breakout board?
12:30 PM CaptHindsight: rs[m]: the one that you mentioned making above
12:31 PM rs[m]: my point is that software step generation is stupid, a concept like putting a pi onto a grbl board as a GUI is stupid, if you could use the µC for step generation, and linuxcnc (or machinekit, whatever fits your bill) to control the machine
12:32 PM rs[m]: for the same amount of hardware
12:33 PM rs[m]: technically, that would also be software step generation, but on dedicated hardware with very predictable timing
12:33 PM CaptHindsight: thank you for letting me know what you think is stupid
12:34 PM rs[m]: sorry
12:34 PM rs[m]: i apologise for being rude
12:35 PM CaptHindsight: I thought a microcontroller generates steps using software :)
12:36 PM rs[m]: but i still don't understand what the relative cost of copper cs. assembled board has to do with anything. that mostly depends on the components on the board and the size of the lot
12:36 PM CaptHindsight: yes it does
12:36 PM rs[m]: s/cs./vs./
12:37 PM rs[m]: yes it does what?
12:40 PM CaptHindsight: somebody has this idea to use STM32's for running all the tasks involved with machine control
12:41 PM CaptHindsight: they just don't know how to interconnect them all or write the application and create all the gui's
12:42 PM rs[m]: "just"
12:42 PM CaptHindsight: yeah, just missing that part
12:43 PM rs[m]: the easy two-weekend-project type of solution would be to get a nucleo and do something like https://github.com/kinsamanka/PICnc-V2
12:43 PM CaptHindsight: long threads on the mail list if you want to help them
12:44 PM rs[m]: i used something like this in a very ad-hoc manner to control a "käsefräse", unfortunately the firmware and the board used are a bit of a mess
12:44 PM skunkworks: software stepping works just fine if you don't need high speed.
12:45 PM rs[m]: software stepping on a rpi with jitter arount +-10000µs and 50000 worst is less than optimal /methinks
12:45 PM rs[m]: s/µs/ns/
12:46 PM CaptHindsight: yet is fine for some of my slow moving operations
12:46 PM rs[m]: and given that you need some kind of breakout board anyways (for the raspberry pi), adding a sub $1 cortex m0 to offload step generation is a no-brainer IMHO (if going the rpi route)
12:48 PM CaptHindsight: the orangi pi has the $1 micro inside the SOC for 38 cents (already included in the cost)
12:49 PM rs[m]: same with beagle bone and the PRUs in the TI SoC
12:50 PM CaptHindsight: the difference betwen the BB and the orange pi is that the orange pi has a GPU that actually works fast enough to be used for a GUI
12:50 PM rs[m]: i really should dig out that old code and put it on github for some forensic programmers to have fun with
12:51 PM CaptHindsight: i think it was 78 cents (already included)
12:51 PM CaptHindsight: too bad machine kit was anti allwinner when they chose the BBB
12:51 PM rs[m]: i don't know the orange pi, i messed around with a tinkerboard and an amlogic s905 (?) i think, but didn't manage to get a preempt-rt kernel to boot and behave
12:53 PM rs[m]: i can't imagine they would mind including allwinner drivers?
12:54 PM CaptHindsight: latency with xenomai on the allwinners was <40,000
12:54 PM rs[m]: the tinkerboard would be very nice, seems to be a much cleaner design than the pi, no (apparent) power issues, available with eMMC, bunch of UARTs and other peripherals, much faster than rpi3
12:55 PM CaptHindsight: somewhere under 100K with preemt_rt
01:00 PM rs[m]: why is "xenomai on allwinner" in the past tense
01:01 PM CaptHindsight: haven't touched it in 6+ years, don't know how the current xenomai will work
01:01 PM CaptHindsight: or on the newer silicon
01:05 PM rs[m]: that http://www.cs.ru.nl/lab/xenomai/raspberrypi.html could be interesting
01:05 PM rs[m]: and that http://www.simplerobot.net/2020/06/picat-xenomai-3-for-raspberry-pi-4.html
01:23 PM rs[m]: CaptHindsight: do you have a pointer to this STM discussion?
01:29 PM CaptHindsight: [Emc-users] Interesting GUI I think was the latest
01:30 PM CaptHindsight: [Emc-users] Machmaker USB -> Parallel
01:31 PM CaptHindsight: [Emc-users] linuxcnc on modern hardware
01:32 PM rs[m]: hmm. interesting gui is about datron neo
01:32 PM CaptHindsight: [Emc-users] Open source CNC architecture
01:33 PM CaptHindsight: Real-time OS for machine controllers
01:33 PM CaptHindsight: rs[m]: yes, started out about that, then changed topics
01:37 PM CaptHindsight: apparently x86 PC's are not modern, and software that works really well and currently in use gets "old"
01:38 PM CaptHindsight: don't know if the bits start to dry out and crack so you can't tell a 1 from an 0 or what actually happens
01:38 PM CaptHindsight: something about style from what i glean
01:38 PM rs[m]: those discussions are best left alone
01:51 PM rs[m]: some things software do get old and "bitrot" or are simply not adequate for newer hardware, (e.g. higher resolution screens)
01:59 PM roycroft: using a modern file system like zfs eliminates bitrot
01:59 PM roycroft: so folks need to find new excuses for brokenness :)
02:29 PM CaptHindsight: roycroft: seems to me to be more about coding style and language chosen
02:30 PM CaptHindsight: and it's has to be new versions of each
02:31 PM CaptHindsight: so an analogy might be having to take an electric uber to work vs your old internal combustion engine car light your grandpa
02:32 PM CaptHindsight: light/like
02:34 PM CaptHindsight: not sure if wearing pants and a shirt is modern enough, you might have to wear an onesy with feet made of hemp, wood or similar modern material
02:43 PM rs[m]: IME experience zfs doesn't get rid of that particular kind of bit rot. if at all it helps you detect it.
03:12 PM Deejay: gn8
05:20 PM unterhaus_: I have some giddings drives, and the installation manual is missing the part about how to set them up
05:20 PM unterhaus_: I guess I'll try the same setup as the previous model and hope it works
06:43 PM CaptHindsight: skunkworks: all you need to do is put that blacklist file in your /rootfs/etc/modprobe.d/
06:44 PM CaptHindsight: and then move it to your home dir when you want GPU working again
06:44 PM CaptHindsight: copy pasta or move it to and from, up to you
06:46 PM CaptHindsight: https://pastebin.pl/view/2f57585f blacklist-gpu.conf
06:46 PM CaptHindsight: skunkworks: ^^ the files needs this name
07:01 PM skunkworks: thanks - I don't need anything else installed for software rendering?
07:19 PM CaptHindsight: skunkworks: just put that file in that directory and reboot
07:19 PM CaptHindsight: skunkworks: when you want to go back to HW rendering then take the file out of that directory and reboot
07:22 PM CaptHindsight: he just said you can name that file anything inthat folder and the kernel modules will be blacklisted, so don't rename the file, be sure to remove it
07:29 PM skunkworks: sure - thanks - will try it this week
08:59 PM Tom_L: if i have 3 encoders in the bit file but only enable 2 in the config= line in the ini does that make the last one GPIO by default?
08:59 PM Tom_L: i just noticed i had 3 in the bit file
09:00 PM Tom_L: instead of configuring them as NullTag
09:01 PM Nyxlayne: hello again
09:27 PM Nyxlayne: I am stuck. I am overcomploicating this somehow. I have a 5i20 based lathe here that has worked in the past. I am trying to get the pncconf program to make a working first stab
09:27 PM Nyxlayne: I need to make the Estop button flip the E
09:30 PM Nyxlayne: stop relay
09:30 PM Nyxlayne: and I'll be able to get started.
09:31 PM Tom_L: net estop-loop iocontrol.0.emc-enable-in <= hm2_[HOSTMOT2](BOARD).0.gpio.xxx.in
09:33 PM Tom_L: i've never used pncconf
09:36 PM Tom_L: https://forum.linuxcnc.org/47-hal-examples/25861-external-e-stop
09:36 PM Tom_L: there's another one
09:45 PM Nyxlayne: i wish I didn't need anything like it, either. Trouble is, I had help setting this thing up and it worked so well that there is a lot that I didn't learn.
09:50 PM Nyxlayne: ls
09:53 PM SpeedEvil is now known as Guest2032
10:38 PM CaptHindsight: skunkworks: if you also get a chance test latency with software rendering and isolcpus=3 and idle=poll
10:38 PM CaptHindsight: vs isolcpus=1,2,3 and idle =poll
10:39 PM CaptHindsight: this should leave cores 0, 1 and 2 free for software rendering and non-RT
10:40 PM CaptHindsight: if rs[m] is right then lowering the max core speed to something that won't cause throttling might also help with latency even thought we are telling the cpu to force it to max core freq using force_turbo=1
10:42 PM CaptHindsight: force_turbo=1 goes into the config.txt file, not the cmdline.txt
11:45 PM CaptHindsight: had some time here, isolcpus=1,2,3 and idle =poll definitely works better than isolcpus=3 and idle=poll
11:45 PM CaptHindsight: have max freq at 1500
11:45 PM CaptHindsight: force_turbo=1
11:46 PM CaptHindsight: 50,000 base thread, 200,000 servo thread, jitter is <50,000 for both threads :)
11:47 PM CaptHindsight: this is with software rendering and watching youtubes (choppy) in FF or Chrome
11:47 PM CaptHindsight: will let it run overnight