Model triggered thermal runaway...
I spent 2hrs debugging this and sorted, should keep a record for others to reference with.
Printer is MK3 upgraded to MK3S (except the filament sensor part).
I was printing ~20mm dia tube structure. Everytime printer reach 11mm height, thermal runaway happened, at around 12mins mark.
Print 2 models on the same run, thermal runaway happened again at exactly the same height, but now 24mins mark (well, 2 models, double the time it need to reach 11mm height)
That's when I figured there is nothing wrong with printer. If something wrong with printer, thermal runaway will happen at the same mins mark, not double the mins mark when 2 models been printed at the same time. Thermal runaway is actually caused by shape of model, or say... the air blowback to hotend caused by shape of model.
Reduce the fan speed to 128-of-256 solve the problem.
So... I guess the shape of the tube cause air to blow back to hotend, and at 11mm height the air is strong enough to overcome the heating capability of hotend, then thermal runaway triggered. And yes, I've E3D silicon sock on, and it still happened...
I'd print the same model with MK3-vertical-fan and MK3-angled-fan before, both no problem. From what I can see, MK3S' fan shroud blow more downward air then others. It's fine until you print something like 20mm tube, when it reach 11mm height.
Hope this would help others
RE: Model triggered thermal runaway...
If you have a sock on this definitely should not be caused by the fan. I've printed all kinds of parts that cause this "blowback" without problem.
Sounds like you have a bad thermistor cable. It is flexing and going open circuit when it hits the 11mm height mark. easiest way to troubleshoot is move Z to 11 (heh) and then exercise the X range of motion and check for the temperature readouts on the display to vary from the current value. If they do, you've confirmed there's a break in the cable. Using a multimeter to do this will likely give a better result since the display doesn't react to changes as quickly as a multimeter in ohms mode can.
RE: Model triggered thermal runaway...
Another check you can do is to run the nozzle PID at the temperatures you are printing. This procedure helps the printer know how to apply power to the nozzle to best maintain an accurate temp. It helps avoid the case where it applies too little power when there is a large heat load.
It may not affect what you are seeing, but then again, it might help.
RE: Model triggered thermal runaway...
I somehow got the same problem again today. And then figure out that this is probably not caused by model causing air blow back on hotend, but a PID problem.
The funny thing is: PID calibration on MK3S actually cant fix the problem. I'd to use M303 gcode to get PID value, and put it in via M301.
If you run into thermal runaway on MK3S. Do M303/M301 PID. Don't use MK3S' PID.
RE: Model triggered thermal runaway...
I somehow got the same problem again today. And then figure out that this is probably not caused by model causing air blow back on hotend, but a PID problem.
The funny thing is: PID calibration on MK3S actually cant fix the problem. I'd to use M303 gcode to get PID value, and put it in via M301.
If you run into thermal runaway on MK3S. Do M303/M301 PID. Don't use MK3S' PID.
IMHO the reason why the PID calibration in the menu of the MK3 line is the extruder PID calibration only, where what you wanted was the bed PID calibration. This is (again IMHO) a failing of the firmware calibration, it should either do both or offer both as separate PID calibration runs.
Thanx for the pointer to the M codes for PID calibration. I may do this on my MK3(upgraded to S) even though I'm not having any issues yet to tune my machine for the MFG tolerances of the components in the PID loop.
See my (limited) designs on:
Printables - https://www.printables.com/@Sembazuru
Thingiverse - https://www.thingiverse.com/Sembazuru/designs
RE: Model triggered thermal runaway...
Argh... I have to retype my edit as the timeout for editing happened while I was writing... F*cking forum setup...
What I wanted to add was I remember seeing a work-around suggested for a similar problem, I think in the conversation for one of the issues over on Github (the bug tracker). The suggestion was to print problematic models off center of the bed. The reasoning given was that because the temperature sensor (thermistor) is mounted to the center of the bed, it would better see fast changes in the bed temperature in the center. But because of the (relatively) slow movement of heat sideways through the bed stackup, rapid thermal changes away from the sensor get "filtered" out before the thermistor can register them. This explanation was presented as a hypothesis without any empirical data to back it up, but it does make sense. (Unless I'm forgetting or not aware of something subtle in thermodynamics. Very possible with my limited understanding of thermo...)
See my (limited) designs on:
Printables - https://www.printables.com/@Sembazuru
Thingiverse - https://www.thingiverse.com/Sembazuru/designs