Printer Instability about klipper HOT 16 CLOSED

Please attach the full /tmp/klippy.log file containing the event and I'll take a look.

from klipper.

Alright!, I'll dry run the same 2 hour print, hopefully it is not random.

from klipper.

from klipper.

Wish I didn't reboot! That cleared the /tmp directory.
Anyways I just noticed that my ttl voltage converter (serial) is not working properly. Instead of disclosing this issue as user error only, hopefully the error code might be a hint for others for communication failure perhaps? Would be good to know.
Thanks.
Edit: sorry for my stupidity, closing until it happens with direct USB connection aswell. Thanks

from klipper.

Hello,
After trying with USB, direct connection to my PC, printer shutdown after 5 hours print 😭 . (Move out of queue)
Here is my klippy log:
klippy.txt
(note: you might find in the log that the printer did shutdown before due to sensor error adc, diy heatbed here, tho firmware/soft restarted and no issues since then. Guess this is completely unrelated)
I hope its something like bad usb cable/software (no dmesg errors for usb), else reliability might be an issue for me.
Thanks.

from klipper.

from klipper.

I simply downloaded the latest master commit as zip 2 days ago on a separate PC, and thus no git.
Here is the gcode:
mod.txt
Edit2: Firmware flashed from the same commit.

from klipper.

Upon further inspection of the log, the drift between the host and mcu clocks sometimes changes rapidly. I suspect this is what led to the above error. I need to further investigate to determine if this is due to some attribute of the AT90USB chip, a code error, or something else.

from klipper.

I'll try replicating the issue on other computer dry run. Could it be host side stuttering due to kernel/hardware issue?

from klipper.

I'll try replicating the issue on other computer dry run. Could it be host side stuttering due to kernel/hardware issue?

Yes, that's certainly possible. If you have an RPi2 or RPi3 that would make an excellent test.

FYI, I've run some tests on my AT90USB1286 chip and I can confirm that the odd timing behavior is not inherent to that micro-controller. I've also identified a few areas in the Klipper clock synchronization code that could be improved. It will likely be sometime next week before that's ready for testing though.

from klipper.

Please try the testing branch work-clocksync-20170627 (git fetch ; git checkout work-clocksync-20170627). I've reworked the clock synchronization code - I think it should be more stable now. It would be great if you could run some prints with the new code and post the klippy.log file (I'd like to see the log even if there is no failures).

Thanks,
-Kevin

from klipper.

Will do with the same setup and gcode uploaded above.
Edit: I have made a 3d printed delta kossel that I will use as main 3d printer with marlin temporarily (will switch to klipper later aswell). Ill keep this printrbot jr for dry run from now on (y axis is broken, sagging due to moisture (wood)), as a testing device for klipper., which is useful for testing stability of future commits/changes) If the following gcode passes, Ill print a 40+ hour job. (with logs and everything, ill also test high print speeds on long prints to see if its host resource issue with very complex prints)

from klipper.

Hello,
I accidentally ran the same klippy branch (didnt remove /etc/default/klippy), same gcode as above, here is the log:
klippy.txt
Printer did not crash. Ill try the same with the latest commit and post the log.

from klipper.

work-clocksync-20170627 branch:
klippy2.txt
No crash aswell. I'll stay on this branch as primary firmware.

from klipper.

Thanks. The log indicates the new code has detected the clock drift. The results should be more stable now. I've committed the changes on the test branch to the master branch.

from klipper.

Thanks, I noticed that its quite repetitive, donno if this has to do with system latency, clock source set to TSC, I hear that its affected with cpu scaling, Ill try switching to HPET, anyways this is host specific now, closing.
Issue resolved.

from klipper.