#linuxcnc-devel Logs

Jul 26 2022

#linuxcnc-devel Calendar

01:37 AM pere: I notice several packages disappeared between the stretch and buster live image: nopaste, f-engrave, truetype-tracer. Were these built locally by linuxcnc, or where did these come from?
04:53 AM silopolis[m]: Hello there, howdy today? 🙂
05:22 AM pere: hi
05:23 AM pere: silopolis[m]: I got a working bookworm live iso yesterday, now I just wait for andy to restructure git to submit the pull request. :)
05:27 AM silopolis[m]: @pere u rock man 👍💪
05:29 AM pere: it feels more like I hit my head on a rock, with the slow movement of the pull requests. :/
05:32 AM silopolis[m]: Understand 😔
05:34 AM silopolis[m]: This is a long standing issue, project now has to face it, take decisions and make moves
05:35 AM pere: silopolis[m]: I believe it is us here that need to face it and make decision, waiting for "the project" is not going to do much good.
05:36 AM silopolis[m]: Starting to think we need a conference by the end of summer with all big boys to really know where and how we're going to 2.9 and after
05:37 AM silopolis[m]: As long as we don't have the keys on GH, we're glued
05:40 AM silopolis[m]: Only option I see is the one @RealThunder chose for it's (amazing) work on FreeCAD : work on its fork and welcome others to contribute, remaining as close as possible from upstream
05:41 AM pere: did you ask for write access to github?
05:43 AM silopolis[m]: Years later, now that FC managers have finally admited his work was the future of the project, he's handling alone the burden of preparing the biggest merge in project history 😢 No wonder he gets my biggest tip on Patreon !
05:44 AM silopolis[m]: It's like if now we'd say "well, ok, in the end you can merge now" to MachineKit ! 😕
05:44 AM silopolis[m]: Which we should btw, if you ask me !
05:44 AM silopolis[m]: pere: Nope, didn't felt legit to do so
05:46 AM silopolis[m]: I'm really not a developper and still feeling I should get a better knowledge of the project and code base for this
05:47 AM silopolis[m]: OTOH, I could use GH credentials to to project management tasks like issues and PRs triage and tagging, settting up a GH Projects test...
05:49 AM silopolis[m]: Maybe I could also review and merge docs related commits and tag Jedis for second review if they're too technical
05:49 AM pere: silopolis[m]: more often than not, such forks just result in fragmentation and lack of critical mass causing the entire project to die...
05:52 AM silopolis[m]: pere: we've both been in FOSS for long enough to know that it's sadly true...
05:54 AM pere: this make me believe it is better to gain the trust of the current gatekeepers and help them until one can transition responsibility to a fresh set of hands.
05:55 AM silopolis[m]: 1000 %
05:59 AM pere: I have been looking a bit at at_pid, trying to figure out which algoritm it is using. unsuccessfully so far. I suspect it is the method described in Automatic Tuning of Simple Regulators with
05:59 AM pere: Specifications on Phase and Amplitude Margins by Karl Johan Åström and
05:59 AM pere: Tore Hägglund (doi:10.1016/0005-1098(84)90014-1),
05:59 AM pere: <URL:https://lup.lub.lu.se/search/ws/files/6340936/8509157.pdf > or
05:59 AM pere: Automatic Tuning of Simple Regulators by Karl Johan Åström and Tore
05:59 AM pere: Hägglund (doi:10.1016/S1474-6670(17)61248-5)
05:59 AM pere: <URL: https://lup.lub.lu.se/search/publication/8601786 >. Anyone here know PID theory enough to know?
06:00 AM pere: I can not get the numbers to quite match up.
06:05 AM silopolis[m]: Man, I'm so illiterate in electronics 😭
06:11 AM pere: this is math, not electronics. :)
06:11 AM silopolis[m]: 🤣
06:11 AM silopolis[m]: Even worth 😅
06:12 AM silopolis[m]: * Even worse 😅
06:17 AM rmu: I don't think at_pid is used much if at all. maybe the original author had an itch to scratch.
06:21 AM pere: rmu: I have no idea if that is the case, but assuming the author (Peter Peter G. Vavaroutsos) was not a control system engineer, I suspect he used some existing and well known algoritm.
06:21 AM pere: s/Peter Peter/Peter/, sorry.
06:27 AM pere: After further study of the code, I am sure it is the Åström/Hägglund algorithm that is used. 0.6/8 = 3/40, so it matches the Ziegler-Nichols step that should be used in the Åström/Hägglund procedure.
06:28 AM pere: <URL: https://en.wikipedia.org/wiki/PID_controller > provide som details on what I am talking about.
06:46 AM rmu: pere: what numbers don't match up`
06:47 AM pere: rmu: the numbers in the Ziegler-Nichols calculations of at_pid.c with the numbers listed in <URL: https://en.wikipedia.org/wiki/PID_controller >. But I find that the C code just have a very convoluted way of doing the same calculations.
06:50 AM rmu: those numbers are somewhat motivated by theory but more or less heuristic AFAIK
07:02 AM pere: could be, but this confirms that the method used is from the paper I pointed to.
07:02 AM rmu: so without looking too closely at the code it seems some oscillation is forced with a +-step set point that is inverted when output changes sign, and that is assumed to be the "critical oszillation period". then "critical gain" is calulcated according to formula (1) in the paper you cited
07:02 AM rmu: that gain is then used to derive P I and D gains according to ziegler-nichols
07:02 AM pere: I believe the output is switched when acceleration is zero, not when output change sign, but yeah.
07:04 AM pere: yeah
07:12 AM rmu: are you going to autotune an analog servo drive with this?
07:13 AM pere: I already did
07:14 AM rmu: and how are results?
07:15 AM pere: better than my unskilled hand-tuned values. :)
07:19 AM rmu: my analog servos suffer from changing zero offset depending on temperature, different overshoot depending on speed, low bandwidth and course encoder resolution -- can't tune that out unfortunately
07:35 AM pere: sound like you need more than a simple pid controller.
07:37 AM rmu: it is usually good enough for what i do, but i could tune it better for specific situations
10:32 AM -!- #linuxcnc-devel mode set to +v by ChanServ
11:08 AM -!- #linuxcnc-devel mode set to +v by ChanServ
11:49 AM -!- #linuxcnc-devel mode set to +v by ChanServ
01:32 PM CaptHindsight[m]: sounds to me like a PID for temp and a PID for motion
01:33 PM CaptHindsight[m]: rmu: does it overshoot more at higher speeds?
01:36 PM CaptHindsight[m]: pere: applying math functions in C is often difficult, you need a dev that understands both
01:46 PM silopolis[m]: Maybe a reason for opencn to have plugged in mathlab?
02:39 PM -!- #linuxcnc-devel mode set to +v by ChanServ
02:55 PM -!- #linuxcnc-devel mode set to +v by ChanServ
03:18 PM -!- #linuxcnc-devel mode set to +v by ChanServ
03:50 PM linuxcnc-build: build #1954 of 4041.deb-buster-rtpreempt-amd64 is complete: Failure [4failed shell_3] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/4041.deb-buster-rtpreempt-amd64/builds/1954 blamelist: Sebastian Kuzminsky <seb@highlab.com>
04:34 PM linuxcnc-build: build #1847 of 4042.deb-buster-rtpreempt-rpi4 is complete: Failure [4failed shell_3] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/4042.deb-buster-rtpreempt-rpi4/builds/1847 blamelist: Sebastian Kuzminsky <seb@highlab.com>
04:44 PM pere: I hope even more patches will make it to master. P)
05:08 PM silopolis[m]: For those interested, here are infos and data I collect about GTC/ISO13399
05:08 PM silopolis[m]: https://drive.google.com/drive/folders/15iHgVWVqBm_SQ9sDjJBDbjBsX17KrE54
06:16 PM -!- #linuxcnc-devel mode set to +v by ChanServ
06:46 PM -!- #linuxcnc-devel mode set to +v by ChanServ
08:11 PM -!- #linuxcnc-devel mode set to +v by ChanServ
08:34 PM Tom_dev: trying to build 2.8.2 from source: File "<string>", line 1, in <module>
08:34 PM Tom_dev: File "hal/user_comps/pyvcp.py", line 49
08:34 PM Tom_dev: SyntaxError: Missing parentheses in call to 'print'.
08:34 PM Tom_dev: not a python programmer... sry
08:39 PM TurBoss: you are trying to use python to build 2.8
08:40 PM TurBoss: ops python3
08:40 PM TurBoss: python --version
08:45 PM Tom_L: no but there are python files in it
08:46 PM Tom_L: i may have it, i'm not sure
08:46 PM Tom_L: indent typo maybe
08:46 PM Tom_L: 2.7.16
08:51 PM Tom_L: ok i got it
09:01 PM Tom_dev: pere, did a clean buster install and tried again
09:01 PM Tom_dev: had trouble installing linuxcnc from package manager so i built from source
09:01 PM Tom_dev: Reading state information...
09:01 PM Tom_dev: E: Unable to locate package linuxcnc-uspace
09:01 PM Tom_dev: E: Unable to locate package linuxcnc-uspace-dev
09:01 PM Tom_dev: E: Unable to locate package linuxcnc-doc-es
09:01 PM Tom_dev: E: Unable to locate package linuxcnc-doc-fr
09:01 PM Tom_dev: E: Unable to locate package linuxcnc-doc-en
09:01 PM Tom_dev: E: Unable to locate package mesaflash
09:01 PM Tom_dev: P: Begin unmounting filesystems...
09:02 PM Tom_dev: also i had to add a new key for linuxcnc which is something new to me
09:02 PM Tom_dev: i used the old key and got an error