Z-calibration question
I had my first "real" failure on the XL this week, and while the failure itself wasn't my question clearly afterwards that extruder (#2) had an altered z-offset (I am assuming it had a collision prior to this failed print since the last material I tried to print was PA-CF, what I don't understand is the internal logic of the XL which is what I'd love to have someone who really knows). One of the greatest features of the MK4 and XL are the glass smooth 1st layers afforded by the load cell that detects that it is on the bed perfectly. Fine, but then I see a print where it seems to fail to squish the filament down (as in the failure) photograph here. So on my CNC milling machine I have a tool setter so when the spindle grabs a new tool from the tool changer it moves over and probes the length so unlike in days of old I don't need to manually measure each tool, and every time the tool is used it gets remeasured, so the offset is always correct (and detects a broken tool if the probe is ~1mm distance from prior). Does the load cell not serve the same purpose except continuous. Regardless that it got knocked, since it probes itself against the bed, how could it ever not have the correct z-height adjustment?
RE: Z-calibration question
I have similar questions, as I've three XLs and I need to slice the same exact model separately with a tiny Z Offset tailored for each machine to get the exact same truly perfect first layer, and I don't think this should be necessary given the calibrations and sensors.
I've some thoughts about the potential issue(s); the pre-print bed sampling only happens with one tool. I presume the full tool head calibration (where you put the pin in so it can learn the exact geometry and placement of each head's nozzle) is then used combined with the pre-print bed sampling data to calculate every tool's exact relative position. But from my experience that is not true; either the calibration/bed leveling isn't precise enough, or there's something missing in the end calculation. I've re-run tool head calibration dozens of times, re-seated nozzles, and lots of other stuff while trying to address this issue (because slicing the same model 3 times and having to keep track which gets sent to each printer is a royal pain) but I've not had a eureka moment yet.
I also have different shaped nozzles in different heads (e.g. a CHT w/ adapter in some, standard brass Prusa nozzles in others) which adds another variable to this, but I don't know why it would prevent the math from working out since that should be accounted for in the tool calibration. Soon I'll be replacing all mismatched nozzles with the new CHT Prusa nozzles to eliminate this as a contributing factor.
It sounds like your tool position calibration is no longer accurate, you may want to re-run that. If it used to be accurate and now isn't, I would guess the nozzle has shifted and you might want to check that they're not loose enough to move around.
Like I said, they theoretically have all the data needed to overcome all of these situations, assuming the tests and sensors are precise enough and repeatably accurate, so I'm not sure what's to blame for the variances. There's also other unlikely-but-possible suspects, like the potential for small amounts of oozing during warm-up introducing load cell inaccuracies during the bed mesh testing.
I'm sure I'm not even considering everything involved, either. Would love other thoughts.
RE: Z-calibration question
From my understanding the bed probing is only done with one tool and the rest of the bed offsets for all other tools is calculated from that one nozzle and the nozzle offset calibration data. I've even seen a couple prints where my nozzle will dip into the purge tower for whatever reason and no crash is detected even with Crash Detection ON. There are no warnings if you don't get a perfect first layer and nozzles digging into the plate from a bad MBL step is still a thing. So, no, I don't/wouldn't believe the loadcell is 'continuously probing.'
There also seems to be a limit (dunno what the actual number is) to how much the XL compensates for a not perfectly tram/flat bed. Otherwise they're wouldn't be complaints about bent beds from bad packaging. If you print out a full single layer sheet you can find your high/low spots of your bed that isn't completely corrected for. I know with a 60c bed that I have a small high spot on one of the top right plates and a slightly low area center-left.
Prusa also added a 'stuck filament detection' feature which I assume uses the loadcell and that gave me so many false-positives whenever I printed PETG that I just turned it off. It's cool tech and can be great, but I wouldn't say we're quite there yet. Personally, I had a perfect experience with the 'MK3+silicone bed mod' and sometimes wish I could revert to that for an actual 'set it and forget it even if nozzle messy' experience.
XL-5T, MK3S MMU3 || GUIDE: How to print with multiple-nozzlesizes do read updated replies || PrusaSlicer Fork with multi-nozzlesize freedom || How Feasible is Printing PETG for PLA supports on XL very
RE: Z-calibration question
Wow, your problem is worse than mine, Tool 4 is working perfectly while tool 2 is too high, But glad it's not just me.
RE: Z-calibration question
Can't learn to fix issues if you don't encounter them ¯\_(ツ)_/¯
Despite my criticisms for perfection, I've had a good time with my XL.
XL-5T, MK3S MMU3 || GUIDE: How to print with multiple-nozzlesizes do read updated replies || PrusaSlicer Fork with multi-nozzlesize freedom || How Feasible is Printing PETG for PLA supports on XL very
RE: Z-calibration question
Me too, my XL up until now has been the most reliable printer I have ever used, with some pretty amazing results
RE: Z-calibration question
It sounds like your tool position calibration is no longer accurate, you may want to re-run that. If it used to be accurate and now isn't, I would guess the nozzle has shifted and you might want to check that they're not loose enough to move around.
If the nozzles were all used at the same rate then their relative values would remain the same - but in my case at least I know that two nozzles are used far more than the others and so must be wearing faster.
I assume that as nozzles wear at different rates in use the full (pin) calibration is no longer accurate and therefore needs to be re-done every few months.
BUT
The commonest reason for a poor first layer is ooze, usually caused by damp filament, causing a mini-pad of plastic to form at the nozzle tip so that the probe values are wrong. Usually it's obvious as you watch the first layer go down and the print can be abandoned and restarted in the first few minutes.
And this in turn is why I regard starting a print remotely by network as a stupid practice.
Cheerio,
RE: Z-calibration question
In my case, the differential between machines is consistently there even immediately after tool head calibration, and I have done several prints where I've cleaned the nozzles before start and then obsessively watched/caught any oozing. I don't get that much oozing anyway, as all five rolls in all three machines are in active drying enclosures (and prints turn out great).
My inter-machine variance is super tiny (0.1mm or less) and I only noticed because of fit issues between machines on some very high precision parts. It may very well be that the precision of the calibration and testing (or the current firmware) just isn't capable of being more precise. But if that's the case, I sure wish there'd be a firmware-base permanent offset so I can use the same sliced file everywhere.
Also other than this issue, I second the reliability and capability of these printers enthusiastically. I've printed thousands of hours across the three XLs and the output quality and success rate from day one has been amazing, even with complex multi-material prints. Sure there're quirks and room for improvement, but that's Prusa's wheelhouse; maturing and improving their products for years. To think this is what the XLs are capable of now, I'm excited to see what comes over the next two or three years.
RE: Z-calibration question
Well after watching years of STNG, pretty sure you have a spatial disturbance there varying the density of space. Did you try using a Heisenberg Compensator? Always worked for the Enterprise!
In my case, the differential between machines is consistently there even immediately after tool head calibration, and I have done several prints where I've cleaned the nozzles before start and then obsessively watched/caught any oozing. I don't get that much oozing anyway, as all five rolls in all three machines are in active drying enclosures (and prints turn out great).
My inter-machine variance is super tiny (0.1mm or less) and I only noticed because of fit issues between machines on some very high precision parts. It may very well be that the precision of the calibration and testing (or the current firmware) just isn't capable of being more precise. But if that's the case, I sure wish there'd be a firmware-base permanent offset so I can use the same sliced file everywhere.
Also other than this issue, I second the reliability and capability of these printers enthusiastically. I've printed thousands of hours across the three XLs and the output quality and success rate from day one has been amazing, even with complex multi-material prints. Sure there're quirks and room for improvement, but that's Prusa's wheelhouse; maturing and improving their products for years. To think this is what the XLs are capable of now, I'm excited to see what comes over the next two or three years.
RE: Z-calibration question
Not yet, so far I had only tried rerouting the calibration routine through the main deflector dish 😀
My inter-machine variance is super tiny (0.1mm or less)
I'd say you are doing well - the base unit of accuracy for an FDM printer is half an extrusion width - 0.225mm for a 0.4mm nozzle by default...
Light tapers are the go-to for tight fitting parts. Regard your prints as blanks - as you would with castings, and machine to fit.
Cheerio,
RE: Z-calibration question
I agree, I'm quite happy with the accuracy and repeatability of these machines. In fact it's this repeatability that highlights the calibration question. Each machine starting Z is off consistently the same amount across hundreds of prints and a dozen tool calibration cycles, and is addressable with live z adjustment or the z offset in the slicer.
We already use tapers, drafts, fit tolerances, etc. as appropriate in designs, and do post-processing and finishing where necessary. The particular feature in my use case here though is reliably resolved with a per machine Z offset, saving a substantial amount of labor and time on a high volume part. Again, this is only possible because the printer exceeds the average (from my limited experience) in desktop FDM accuracy/precision, so it's all good!
One would imagine, given the stated precision of the load cell and accuracy of CNC-like tool touch probes being at or better than 0.01, we could achieve even better on that initial Z. If not, I can live with that; but give us a firmware based permanent offset, like on previous Prusa printers. There's already the live Z adjustment, so the heavy lifting is already being done printer side. This new value would just be the default starting point for the live offset.
I'd say you are doing well - the base unit of accuracy for an FDM printer is half an extrusion width - 0.225mm for a 0.4mm nozzle by default...
Light tapers are the go-to for tight fitting parts. Regard your prints as blanks - as you would with castings, and machine to fit.
Cheerio,
RE: Z-calibration question
Fate sent me a signal; in searching for a totally unrelated calibration link, I found this regarding fine-tuning of the auto tool head calibration: https://help.prusa3d.com/article/multi-tool-manual-calibration-xl_470560
No reason I shouldn't be able to dial in my observed Z variance here, and solve my issue. Sorry it doesn't directly do anything for OP's issue though (and apologies for sorta hijacking your original issue).
RE: Z-calibration question
Ok, so final result. I did the whole dock calibration, and it failed on tool 5 (previously perfect) and then threw a “unable to perform precision movement” error and to check the belts. I ran the belt tuner site (kind of clever solution) and the right side was dead middle the left side took 2.5 turn to loosen (120hz!). It is very rare for a screw under tension to tighten due to vibration. But all is right with the world it seems.