I suggest you do a print that you have done before that you know works well. My theory is a problem with the z axis in some way. Go around with a spanner and make sure everything is tight, including the couplers. Next check the wiring, even go so far as you unplug the z cables from the Rumba and reinstall. Also check that it runs freely, move it all the way to the bottom and back up again, that can be part of the 100mm up/down test, change it to 300mm. Also try swapping the stepper drivers around on the Rumba too. It might be a dodgy connection on the pins. Also make sure the electronics cooling fan is running and running at full speed and not impeded by detritus.
Machine off. Swap E0 and Z stepper drivers. Machine on. Move X and Y and confirm motion on Z indicating mesh levelling active. Move Z from 0 to 100 three times. Confirm positions are the same each time. Print ZBed alignment print Print first small item - good Print large item with a M600 filament change ... in progress ...
OK. Failed again. More data this time. The first three layers printed fine, I think. Note about these first three layers, hardly any retraction. Swap before starting layer 4. Pattern begins. The initial connection looks fine. Before this layer even completes, the head is way higher than it should be and the print is dead. Killed print. Measure height of nozzle - 6mm to the bottom of the fan shroud Now repeatedly move the bed from 0 to 1 mm. In each case, 3mm exactly to the bottom of the fan shroud. So here we have two motors failing in tandem (unlikely) or a problematic stepper driver (but I swapped that) or a problem on the RUMBA (feels unlikely) or something in the slicer (though surely if it's asking the head to go to 1mm and it isn't, can we blame the slicer? I've looked at the starting script and there's nothing crazy in there like a Z step change. LCD check says Z steps is 1600. The thing I can definitely say is that this print as worked in the past with 0.3 layer height. Let's just get that mentioned ... Here I moved to 0.25 layer height, I didn't change the retraction vertical lift (which I think was what Greg meant by Z-hop) so that's still at 0.3. Bear in mind however that the beginning of this dive into catastrophe was with 0.3 layer and 0.3 lift, which had been working since I built the machine. What can I think of that might have changed. Octoprint had a new release, and possibly that was just before the problems. But that should just be passing GCODE from a to b and I'm pretty sure everyone else would have had similar issues. There are a lot of retractions and lifts/drops on this layer and it feels like this is the issue. Z motion is somehow not doing precisely what it should. Next step will be to check the couplers as suggested by Greg. Removing and repluggin the wires on the rumba will be a bugger as they're screwed in on the back terminals, but I did make a good job of that initially, and as I say, it's been good for months. That'll be the last thing I try as it'll require the base off. Cooling fan checked. No issues. It gets regular hoovering in any case and the RUMBA is free of dust and clean as you like.
What material are you using? Can I see photos of the failed print please. It won't be Octoprint, it only send the g-code to the printer and has no g-code injection functions. I put my money on the wiring. The symptoms are mechanical but i expect the actual fault is electrical.
Thanks for following this Greg; I really appreciate your input. My first thought on the problem this morning is that, if the problem is a Z motion problem, why don't I boil down the print to just the z moves. This will be large list of moves corresponding to each retraction/move step. Notepad++ together with a plugin called LineFilter 2 let me very quickly get to this data. I preceded all the move steps with an autohome and move to center bed. What then follows is about 815 move steps in the Z direction. The file I ended up with is attached. Essentially you can see here the 0.25 layer height and the 0.3 retraction vertical lift. I ran the test three times with identical results each time. The Z height ends up with 2mm, with 2 shown on the LCD, and a measured height to the bottom of the fan shroud of 4mm. In other words, this set of steps does not produce the problem if we boil it down purely to a set of Z steps. I'll next add an M600 in the middle, just in case there's something about the way that is implemented that is causing it. I haven't yet re-tightened the grub screws or altered the wiring. I'd really like to pinpoint this issue so I don't want to change too much at a time. I can say that the grub screws feel solid and there's absolutely no play. I've also added a couple of pictures of the failed print. First while it's in motion: Filaments being used here are E3D Everyday PLA black and white. The white goes down first, 3 layers at 0.75. All fine. Above you're seeing layer 4 being printed, which is the pattern. This is a single layer above, but you can just about make out the front edge where the black has been laid perfectly on the white and the head height, which is clearly at this point (toward the end of the layer) MASSIVELY more than 0.25mm above the model. In fact, as the layer progressed, the head moved further and further away. This got me to thinking about whether it's a real number manipulation issue, but again, I've printed this model successfully before, so it's some sort of failure that occurred within the last two weeks. What follows are two pictures of the mess after it's cooled and removed. The permiter around the edge looks OK. The inside inner and outer perimeters look OK, but to me look as if they're a little thin, the beginning of the issue. Then as the infill begins at the bottom right, the head is already too high (together with the relatively high infill speed) for the infill to not even make it to the perimiter. By the time the print has progress to the top left, the head is so far above the model that it's little better then squeezing a tube of toothpaste onto the floor.
turn off z-hop, I don't think there is a regular failure in the motion of Z. I think there are random missed-steps cause by electrical faults, by removing the z-hop you massively reduce the amount of movements on that axis and should effectively not see the error. the solid layers will have less retractions and z-hop movements so they will look better. The infill layers will have more retractions and therefore greater z-hops which will amplify the error, so it will look worse on those layers. That being the case you need to check the wiring, or the stepper drivers. There could be a fault at either end of the cables.
OK. I'll definitely print one off using no z-hop. It makes sense that this will work. If there's a problem with Z-hopping though, shouldn't I have isolated and amplified that with the Z-Motion-Test.gcode file? At this point I'll give anything a try anyway as it's just a large square box that does nothing useful for me. I've already begin clearing space so I can get the bottom off and rewire the entire RUMBA board. I'll run a no-hop print first as confirmation.
No-hop print went fine. Complete rewiring troublesome. As my original wires were straight into the screw-down connectors, I actually soldered them originally to create a better connection when they go in. e.g. here with the green. Those damned small screws though! The only thing I had that fits was the ceramic screwdriver, which I sadly broke off while undoing the screw for the red wires. I couldn't find anything else (yet) to redo that, but did the blue, black and green. I've also unplugged and replugged all the other stepper motor connections and anything else that can be easily removed and re-seated, including the stepper motor connects onto the Z motor themselves (which were not easy). I've set another print of the original model going that will use z-hop to see whether anything here has helped so far. If not, it'll be time to get the bottom off and a screwdriver that fits and properly get everything off and back on (and probably take the time to label everything while I'm at it).
Whilst this makes it easier to fit it is prone to failure in the longer term, because solder gradually squeezes out from between the twisted wires so the joint loosens. With a simple twisted joint, the copper hardens as the clamp is screwed down so the joint remains firm. You can get crimped ends to use under screw clamps but that's over-kill. Greg's comments on Z-hop is interesting and raises the question as to whether the motors are able to respond quickly enough to the signal and that may be attributed to the Z axis speed. I use 0.25 mm lift and Z speed is 6mm/s on the early lead-screw and have no issues with large area printing.
So ... the z-hop print still fails, and it's identically happening on print layer 4, after the layer change, this time with the filaments swapped over: Actually I did look at the solder and wondered if I'd get a comment like this Mike; i.e. great idea but just don't. It stems from thinking it'll be better whereas clearly experience disagrees. As you can see on in my z-motion-test.gcode, my speed is also 360, or 6mm/s. I can only keep rabitting on about how I also never had a problem until a couple of weeks ago. I'm in agreement to rule out Octoprint upgrade as that's just gcode transmission. Gcode on the other hand looks fine, but I think there was an S3D upgrade not that long back too. It's bloody weird this. The only comforting thing is that it's consistent, so I shall be absolutely clear on when it's resolved. I feel like I'm looking down the barrel of a full rewire on the RUMBA. At the very least, pull those wires out, snip the solder, and try to cram the twisted wires into the teeny hole (from an awkward angle).
Well, I've run out of ideas for the moment and I hate a problem that I can't pin down, so I feel your frustration. I am sure you'll keep us up-to-date on this one!
Status. Switched to RC8 build from Pipshag. Also turned on LIN_ADVANCE, dropping BaudRate to 115400. Still reproduced the issue, thus ruling out Marlin. Since then have: Removed and replaced every single wire from the Rumba. Removed solder from Z wires connecting to RUMBA, revealed and twisted new cable and re-screwed in. Removed and cleaned corner rods and lead screws. Refitted all Tested So far, printed a small print with Z-hop on - went well Printed a set of 4 bag clips with Z-hop on - went well I have to re-print the model with the problem before I can claim success ...
It feels like this problem is solved following the rebuild. I just successfully printed a 120mm twirl, using 0.3mm layer with z-hop on for every move. The print looks flawless. Thanks to all that offered thoughts and guidance.
As I said, I hate problems that go away without a definite issue being found. Good luck with your efforts.
Well, removing and re-applying all the connections seems to have solved the issue. I'd say the problem was most likely a loose connection, and very possibly your comment which led me to remove the solder on the z-motor wiring was the actual solution. Without going through an extremely length process to treat each connection individually and re-test on every connect, I don't think you could actually point the finger at a specific connection. On the other hand, the process of re-doing it seems to have worked, and it wasn't too arduous. If anything, I got something out of it, it being the first time I'd removed the bed since build (to give me a little more space), and I can see now how simple it would be to refit Mitsumi bearings on the bed corner rods should I ever take that route.