#linuxcnc Logs
Feb 19 2023
#linuxcnc Calendar
02:04 AM Deejay: moin
02:14 AM Loetmichel: roycroft: 2.5 hundrets of a mm off over close to 4 meters of travel sounds pretty good. wouldn't call that "crap".
03:33 AM beef[m]: I don't have time myself for this right now, because I'm in the middle of moving my dental laboratory into a new facility this weekend (have to be ready to go by monday morning) but if you don't figure something out here you can probably also check for help on the PrintNC discord under the LinuxCNC channel, since it'll probably be similar setup.
03:35 AM beef[m]: did you find any answers in regards to this by the way?
03:48 AM travis_farmer[m]: G'Morning ☕️
03:59 AM SimplyFast[m]: HI Beef, no I have not got any feedback in regard with REsonance Compensation in LinuxCNC
04:01 AM jpa-: one could probably adapt the klipper input shapers to linuxcnc HAL components https://www.klipper3d.org/Resonance_Compensation.html
04:02 AM jpa-: or possibly the already existing "biquad" HAL component could work if you can figure out suitable parameters
04:08 AM jpa-: https://github.com/Klipper3d/klipper/blob/aa481fd9c3fc915537a51966320773fcf7008e7b/klippy/extras/shaper_defs.py#L33-L44 https://github.com/Klipper3d/klipper/blob/54e21a7c26a65fa75f3b072e5f02d277321453ca/klippy/chelper/kin_shaper.c#L91-L103 apparently more like FIR filter - should be quite simple to implement as a component
05:02 AM rmu: resonance compensation reminds me of the discussion of record players and catastrophic resonance in gödel escher bach. probably also applies to 3d printers, i.e. there is no general solution to avoid catastrophic resonance in all cases.
05:15 AM jpa-: catastrophic resonance is avoidable if the printer is not ridicuously flimsy in the first place
05:16 AM jpa-: after all one can always slow down acceleration
05:17 AM rmu: what about jerk-limiting movement? wouldn't that help?
05:18 AM jpa-: definitely
05:18 AM jpa-: if linuxcnc only did that ;)
05:18 AM rmu: there is 3d printer software that does it
05:18 AM rmu: but i have no idea how effective it is
05:19 AM sensille: i would be interested in working on that at some point
05:21 AM rmu: https://github.com/synthetos/TinyG/wiki/Jerk-Controlled-Motion-Explained
05:29 AM JT[m]: Morning
05:58 AM Tom_L: morning
05:59 AM Tom_L: JT[m], make any progress?
06:01 AM Tom_L: https://www.reddit.com/r/Starlink/comments/wa4kwv/poor_cable_connection_issue/
06:01 AM Tom_L: https://i0.wp.com/olegkutkov.me/wp-content/uploads/2022/11/starlink_repair_nov_4_2022_bad_spx_connector.jpg
06:01 AM Tom_L: i'd double check the cable...
06:01 AM Tom_L: do not hot plug it!
06:13 AM JT[m]: No progress
06:14 AM saquzi: i have 7i97+7i85, cables and power connected, the 7i97 i have flashed with the proper bitfile, but when I type in the command line: sudo mesaflash --addr 10.10.10.10 --device 7i97 --readhmid it wont show nothing for the additional 7i85 sserial or encoder inputs?
06:14 AM saquzi: seems to show everything for the 7i97 though
06:14 AM Tom_L: JT[m], check those links
06:15 AM Tom_L: saquzi, it may not show the sserial ones
06:15 AM JT[m]: Reddit wants me to install an app
06:15 AM Tom_L: check the cable ends very close
06:16 AM JT[m]: Ok
06:16 AM Tom_L: for corrosion or burns
06:16 AM Tom_L: look at that last pic
06:16 AM Tom_L: but power down furst
06:16 AM Tom_L: first
06:17 AM travis_farmer[m]: i like "furst" better ;-)
06:17 AM Tom_L: phoenetics
06:18 AM travis_farmer[m]: huked on fonix werked for me ;-)
06:21 AM Tom_L: saquzi, or your sserial isn't working right.. double check the wiring
06:21 AM travis_farmer[m]: and that a sserial bit file was flashed to the mesa board
06:21 AM JT[m]: Use --dbname1
06:23 AM JT[m]: Or in a terminal
06:23 AM JT[m]: Halrun
06:23 AM JT[m]: LoadRunner hostmit2
06:24 AM JT[m]: LoadRunner hm2_eth board_ip 10.10.10.10
06:24 AM JT[m]: Stupid spell checker
06:24 AM JT[m]: It's loadrt
06:25 AM Tom_L: JT[m], yesterday's log starting around 9:45
06:25 AM JT[m]: In another terminal
06:25 AM JT[m]: halcmd show all hmm > pins.txt
06:34 AM saquzi: all of the same leds light up on the 8i20 no matter do i plug it now on the 7i97 vs 7i85 serial 0..
06:35 AM saquzi: had the one 8i20 working earlier only with the 7i97, now its time for the rest.
06:42 AM saquzi31: should i see p2 or p3 too? Now only p1 what seems to be "empty"
06:42 AM saquzi31: IO Connections for P1
06:42 AM saquzi31: Pin# I/O Pri. func Sec. func Chan Pin func Pin Dir
06:42 AM saquzi31: P1-01/DB25-01 34 IOPort None
06:42 AM saquzi31: P1-02/DB25-14 35 IOPort None
06:42 AM saquzi31: P1-03/DB25-02 36 IOPort None
06:42 AM saquzi31: P1-04/DB25-15 37 IOPort None
06:42 AM saquzi31: P1-05/DB25-03 38 IOPort None
06:42 AM saquzi31: P1-06/DB25-16 39 IOPort None
06:42 AM saquzi31: P1-07/DB25-04 40 IOPort None
06:42 AM saquzi31: P1-08/DB25-17 41 IOPort None
06:42 AM saquzi31: P1-09/DB25-05 42 IOPort None
06:42 AM saquzi31: P1-11/DB25-06 43 IOPort None
06:42 AM saquzi31: P1-13/DB25-07 44 IOPort None
06:42 AM saquzi31: P1-15/DB25-08 45 IOPort None
06:42 AM saquzi31: P1-17/DB25-09 46 IOPort None
06:42 AM saquzi31: P1-19/DB25-10 47 IOPort None
06:42 AM saquzi31: P1-21/DB25-11 48 IOPort None
06:43 AM saquzi31: https://forum.linuxcnc.org/27-driver-boards/48260-7i97-7i85-bit-file this is the bitfile (and only thing) i loaded to the 7i97
07:27 AM Loetmichel: *BOOOARPSSSS*... that was a nice birthday meal. a "Salamanca" burger with 4 patties, fries, a tuna wrap and 12 mozzarella sticks. Why am i even wondering that i am at 106kg? :)
07:58 AM travis_farmer[m]: only 106kg? 136kg here, and i eat far less than that...
08:12 AM sensille: i have a patch that adds a stepdown and depth to the facemilling tab of qtdragon. would it be worth to submit it, or is it of no interest to anybody else?
08:35 AM JT[m]: Tom_L checked cable connections under magnification and they are perfect.
08:40 AM ZincBoy[CAON][m]: This is not currently possible as lcnc only has a second order motion planner. To implement something like the motion shaping you need a third or higher order planner to limit jerk, snap, crackle, and pop. I would be possible to filter the gcode before sending it to lcnc but that breaks if you are using any real time inputs.
08:49 AM unterhaus_: it's not that hard to mail a small padded envelope to the U.K., is it?
08:51 AM unterhaus_: they say 1st class is for documents only, seems strange
09:27 AM Tom_L: JT[m], same error message?
09:28 AM Tom_L: probably time to call them
09:28 AM SimplyFast[m]: Yes, the goal is not to calculae resonance frequency during a cut, its to measure the axis as loaded, pre machinining so that you can get better tolerances and and finish specs by either avoiding the excitation or actively phasing the pulses to try and kill the resonances etc..
09:29 AM SimplyFast[m]: But ok we miss the code to do higher order planning
09:30 AM ZincBoy[CAON][m]: Yes, that is what a higher order motion planner does. With a property tuned 3rd or higher order motion planner you do not need to do input shaping. Input shaping is faking a high order planner.
09:30 AM SimplyFast[m]: This would particuralely help CNC Routers for example, that can have high-ish travel speeds but loack a bit in structural stiffness
09:30 AM ZincBoy[CAON][m]: Or my machine with a 1.5ton y/z axis 🙂
09:30 AM SimplyFast[m]: bhahbhab
09:30 AM SimplyFast[m]: yes
09:30 AM c101horse8332[m]: afternoon yall, back with more issues
09:31 AM c101horse8332[m]: what am i doing wrong this time?
09:31 AM SimplyFast[m]: Would this 3rd order planner be a collosal feature request? I was thinking simple parameters derived from offline axis testing.
09:32 AM ZincBoy[CAON][m]: Yes. 3d printers can do it as they only really need it for 2axis linear moves.
09:32 AM SimplyFast[m]: I'm not a Dev so that's why I ask, I do have a lot of experience as Functional Analyst and Architecture but I learned to ask the person who actually has to do it before declaring anything feasible lol
09:33 AM ZincBoy[CAON][m]: When you start to look at 9 axis moves and real time planning it gets really hard.
09:33 AM SimplyFast[m]: My router mainly works in 2d, I ws thinking man I need that code form my printer
09:33 AM ZincBoy[CAON][m]: Lots of phd level research still going on in this space.
09:33 AM SimplyFast[m]: but in the router space it would be a lot less work
09:33 AM SimplyFast[m]: limit the function to XY as release 1
09:34 AM SimplyFast[m]: cover the majority of real world use cases
09:34 AM SimplyFast[m]: Yes once you hit the 5th axis and on its wonky af
09:34 AM ZincBoy[CAON][m]: It is still not super easy. I would think you would need at least 3 axis to be useful for most lcnc applications.
09:34 AM SimplyFast[m]: lol
09:35 AM SimplyFast[m]: Yes I was also thinking XYZ the easier setup
09:35 AM SimplyFast[m]: but not A or B or C or...
09:35 AM ZincBoy[CAON][m]: The issue is that lcnc is a real time system so you need to be able to re-plan the forward looking motion each servo period.
09:35 AM ZincBoy[CAON][m]: The lookahead motion planner is lcnc is already only 3 axis.
09:36 AM SimplyFast[m]: hmmm, now this is interesting, printer uses a batch comm with timestamped events
09:36 AM SimplyFast[m]: because its slow ass not RTOS
09:36 AM ZincBoy[CAON][m]: Yes, the printers (klipper) use a streamed set of commands.
09:37 AM SimplyFast[m]: my simple mind would be to do 'fixed frequency shaping' based on offline testing of axies, not real time while cutting
09:37 AM ZincBoy[CAON][m]: This does not allow for real time overrides as needed in most lcnc applications.
09:37 AM SimplyFast[m]: like you apply corrections and strategy but don't retune it closed-loop in real time
09:38 AM ZincBoy[CAON][m]: That is exactly what a high order planner does. You define your jerk/snap/crackle/pop limits and don't change them.
09:38 AM SimplyFast[m]: I just saw what it did for my printer and was like wow I want that for my router
09:38 AM ZincBoy[CAON][m]: Tuning is done off line
09:38 AM SimplyFast[m]: but the router runs lcnc
09:39 AM ZincBoy[CAON][m]: 100% agreed that it would be great in lcnc. My machine is really limited due to infinite jerk.
09:39 AM SimplyFast[m]: it may be all moot due to the cutting forces applied etc...not sure...or some of the resonance may excite the cutter too....chicken or the egg? I like omelets!
09:40 AM SimplyFast[m]: If LCNC has API it could callout to Klipper's resonance frequency function?
09:40 AM SimplyFast[m]: now I"m probably just dreamin'
09:40 AM ZincBoy[CAON][m]: Cutting forces/resonances can be mitigated with toolpath strategies. Motion artifacts can't be compensated so easily.
09:41 AM ZincBoy[CAON][m]: The motion planner is integral to the inner real-time parts of lcnc. It is a module but is solidly in the "here be dragons" realm of coding.
09:41 AM SimplyFast[m]: I wonder if we could extend LCNC to Klipper thru API...
09:41 AM SimplyFast[m]: hmmm...
09:42 AM SimplyFast[m]: is that like burning the LCNC bible? I"m new to forum don't want to say things that will get me banned lol
09:43 AM ZincBoy[CAON][m]: No, this is something that has come up many times. There have been a few attempts at it but the consensus is that it probably won't happen until someone with deep pockets pays to have it done.
09:43 AM ZincBoy[CAON][m]: There are some threads in the forum if you look around.
09:43 AM SimplyFast[m]: The API or the Resonance Frequency Comp?
09:44 AM ZincBoy[CAON][m]: Higher order motion planner. There is an API but it doesn't really bypass the motion planner.
09:44 AM SimplyFast[m]: Gotcha
09:44 AM SimplyFast[m]: So higher order motion planner is a lot of effort $$
09:45 AM SimplyFast[m]: API doesn't bypass the motion planner, right needed to have those offset pulses phases
09:46 AM SimplyFast[m]: I wonder how Klipper got it done so fast that RepRap also jumped and did it and now Bambulab did their own as well, probably commercial interests like you said
09:46 AM CloudEvil: SimplyFast[m]: It's a case of 'you don't want to start from there'.
09:47 AM SimplyFast[m]: Its probably less clean under the hood than I think too lol
09:47 AM ZincBoy[CAON][m]: 2d offline motion planner is relatively trivial. Plus lack of need for compatibility with other systems.
09:47 AM CloudEvil: SimplyFast[m]: The architecture is all wrong for it. If you plan it (or your design happens to) then it gets a lot easier to implement
09:47 AM SimplyFast[m]: Yes, LCNC is agnostic
09:47 AM CloudEvil: And that. Offline is mostly cheating
09:47 AM SimplyFast[m]: But it should do the 2d offline module to help the router guys ahahah
09:48 AM SimplyFast[m]: Learn before climbing evrest 9axis robot arm
09:48 AM SimplyFast[m]: Speaking of I'm eyeing the Igus 6 axis arm
09:48 AM SimplyFast[m]: they have a low cost offer now
09:49 AM SimplyFast[m]: I'm in Ottawa btw I see CA, ON
09:49 AM pcw--home: There already is an userland TP fork of LinuxCNC
09:49 AM SimplyFast[m]: well I'm in Gatineau, QC
09:50 AM pcw--home: or maybe more than one (and they support limited jerk)
09:50 AM SimplyFast[m]: I would offer my time as tester and whatever is needed, but I'm no Dev that's the hick
09:51 AM SimplyFast[m]: It's too good to not have it on the noodle router ahah
09:55 AM rmu: pcw--home: you mean the swiss/french one?
09:56 AM rmu: https://opencn.heig-vd.ch/
09:57 AM pcw--home: Yeah, and also there's someone on the forum with a different limited jerk non-realtime TP
09:57 AM pcw--home: Troy?
09:58 AM pcw--home: I Think the recent forum one is closer to LinuxCNC
10:00 AM rmu: opencn is running on a xenomai kernel with a 10khz loop IIRC, and it seems they are building it for rpi4 also, that would be interesting for experiments
10:01 AM rmu: but also AFAIR opencn only works with ethercat at the moment
10:02 AM pcw--home: They do use hal though
10:30 AM * travis_farmer[m] gets ready to go out to eat with family, at a semi-fancy steak-house :-)
10:31 AM Tom_L: had Tx roadhouse last evening
10:32 AM travis_farmer[m]: this is Mac's Grill, in Minot, Maine :-)
10:32 AM Tom_L: ^^ call ahead or wait 2 hrs
10:32 AM Tom_L: no matter when
10:33 AM travis_farmer[m]: yeah, we have reservations already :-)
10:34 AM Tom_L: sea food should be good in your area
10:34 AM Tom_L: i'm sure JT will say it's better in Alaska which i wouldn't argue
10:35 AM travis_farmer[m]: yes, but we go direct to the coast for that. eating lobster while watching the lobster boats come in just has a level of charm ;-)
11:00 AM JT[m]: Looks like I need to find ar5523 driver for the USB wifi adapter
11:01 AM unterhausen: msft doesn't like me blocking cookies
11:04 AM Tom_L: JT[m], windows or linux?
11:05 AM JT[m]: Debian
11:05 AM JT[m]: Gotta run
11:07 AM JT[m]: Bookworm
11:08 AM Tom_L: https://packages.debian.org/bullseye/firmware-atheros
11:09 AM Tom_L: https://packages.debian.org/bookworm/firmware-atheros
11:42 AM DaveB[m]: Can anyone point me in the right direction for finding out, or a means of testing, the polling rate for a userspace component. I know the docs/forums etc say it's not fast/often, but I'd like to have an idea if it'll be 1hz, 500hz, or somewhere in between. It's a pretty big range.
11:42 AM jpa-: DaveB[m]: is it a userspace component you made yourself, or something premade?
11:43 AM jpa-: you could modify it to toggle an output every time its poll loop runs, and halscope that
11:44 AM jpa-: 10Hz pretty reliable even from userspace, most of the time you can get 100Hz but it might miss a loop or two if the system gets loaded down
11:59 AM perry_j1987: there we go got a fresh install of debian 12
11:59 AM perry_j1987: and linuxcnc installed
12:04 PM perry_j1987: Lcvette[m] you around
12:05 PM JT[m]: Tom_L I don't see the deb
12:06 PM Tom_L: that bookworm link is 404 now
12:06 PM Tom_L: must have just updated
12:06 PM JT[m]: Was there a deb
12:07 PM Tom_L: i didn't look
12:07 PM perry_j1987: https://github.com/Lcvette/qtpyvcp-bookworm-installer when its going through the script i just noticed it say cd: qtpyvcp/: No such file or directory
12:08 PM JT[m]: I think I'll have to get internet to get it
12:09 PM Tom_L: if you can find it i can dl and post it for you
12:09 PM Tom_L: then pull the file off the phone
12:14 PM Tom_L: looks like a kernel module
12:14 PM Tom_L: whatever that means
12:15 PM Tom_L: https://debian.pkgs.org/10/debian-nonfree-arm64/firmware-atheros_20190114-2_all.deb.html
12:15 PM Tom_L: maybe
12:18 PM Tom_L: aarg, that's arm64
12:19 PM Tom_L: https://debian.pkgs.org/11/debian-nonfree-amd64/firmware-atheros_20210315-3_all.deb.html
12:20 PM Tom_L: nothing for sid yet
12:28 PM Tom_L: https://packages.debian.org/bullseye/firmware-atheros
12:28 PM Tom_L: there's some zip files there on the right
12:29 PM Tom_L: not exactly sure what i'm looking for
12:52 PM bjorkint0sh: https://www.youtube.com/watch?v=YR5duZwe1VA
01:49 PM JT[m]: Tom_L I got the laptop to pass through once, I think it has to be pc to pc with no switch
01:53 PM JT[m]: Tom_L I connected to the rpi like that
01:55 PM ZincBoy[CAON][m]: It really depends. The inputs and outputs to the user space comp will be updated every thread period. The actual shared memory object operates a bit like an asynchrous fifo and there is no defined timing so an update rate is not really the right way to look at it. It is more like there is an undefined latency between the uspace and realtime layers. I just make sure to have synchronization logic when doing a bidirectional
01:55 PM ZincBoy[CAON][m]: communication.
02:55 PM CaptHindsight[m]: doesn't Tormach have a 9-axis motion planner for it's version of machinekit/LCNC?
02:57 PM CaptHindsight[m]: and wasn't it Tormach that extended the 3-axis motion planned that is currently in LCNC to 50-100 step look ahead from the previous 2 or so?
03:52 PM cradek: linuxcnc has had the new planner developed at tormach, for years
03:53 PM cradek: I'm so far behind I'm just starting to use it, but it seems great
04:02 PM Total[m] is now known as Total0858[m]
04:06 PM travis_farmer[m]: JT: let me get this right, you can connect your laptop to your phone, and share the phones internet connections to one down-stream device via ethernet? couldn't you just plug that laptops ethernet into the WAN input on your router? that should share the internet connection to the rest of your network. may take some router tweaking, but it should work, in theory.
04:16 PM Tom_L: sounds like alot to have to undo
04:17 PM travis_farmer[m]: could be, but it's a temporary solution
04:20 PM travis_farmer[m]: obviously, if his regular internet worked, it would be far simpler. i have in the past used an old router in place of the current router, and plug a WiFi repeater into the WAN of the router, then connect the WiFi repeater to the cellphone WiFi hotspot
04:37 PM JT[m]: Didn't work when I tried to use the lan
04:42 PM JT[m]: I rebooted the pc and was able to connect to my phone I think
04:42 PM -!- #linuxcnc mode set to +v by ChanServ
04:42 PM Tom_L: seems to work
04:43 PM JT-Shop: yup... finally
04:43 PM JT-Shop: now to get starlink to work lol
04:43 PM Tom_L: what is this going to accomplish?
04:43 PM Tom_L: heh
04:43 PM JT-Shop: anyway I can work now
04:44 PM JT-Shop: I can still work and ship Mesa cards
04:44 PM Tom_L: true
04:44 PM JT-Shop: more important get my Leo fix
04:45 PM JT-Shop: I do need to print from the laptop to the laser printer to make shipping labels...
04:45 PM Tom_L: is the printer wifi?
04:46 PM JT-Shop: yea
04:46 PM Tom_L: that should be fun
04:46 PM JT-Shop: but I'm on the phone wifi...
04:46 PM Tom_L: took me a while to reset mine once
04:47 PM Tom_L: no usb to it?
04:47 PM Tom_L: likely easier to do that than wifi
04:47 PM Tom_L: i _think_ mine has all 3
04:47 PM Tom_L: wifi, usb & lan
04:48 PM JT-Shop: just looked and one has usb port so maybe I can print to it
04:49 PM JT-Shop: the main one is wifi only afaict
04:50 PM Tom_L: probably have to edit or reinstall a driver
04:57 PM JT-Shop: deleted the printer then plugged it in and like magic it worked...
04:58 PM Tom_L: about time something did
04:59 PM XXCoder: printers ugh
05:00 PM JT-Shop: Tom_L, you got that right
05:02 PM JT-Shop: at least I have a backup now that can be deployed easy now that I know how...
05:32 PM roycroft: https://www.youtube.com/watch?v=ckIMuvumYrg
05:35 PM XXCoder: hah dont need captions to be funny
06:11 PM XXCoder: .
06:12 PM TurBoss8676[m]: hurray!
06:13 PM XXCoder: yeah finally. missing user error on bot
08:32 PM _unreal_: very long day today
08:32 PM _unreal_: went to the beach btw
08:58 PM Guest16: Has anyone seen a case when Raspberry Pi4 fills the disk with 'received approval for ...' from the dhcp server due to a constantly broadcasting gratuitous arp requests?.
09:13 PM Tom_L: nope
09:44 PM CaptHindsight[m]: cradek: I thought that I read about a newer trajectory planner than the current latest in LCNC that you mention