I am having the same issue with P1S, consistently undersized by about 0.18mm in XY. The same G-code produces accurate prints on a non-bamboo printer, so it is not a problem with the BambuStudio slicer. It’s likely a bug with the firmware, introduced when implementing some smarter features. I have to scale thing up in CAD, as PEZ3D says, slicer would no correct it. I hope someone can formally report this to Bambulab.
1 Like
The last time I checked, the prusa firmware was open source. I know this because in the past I was able to make some minor changes to it to compensate for the added height of a volcano hot-end, recompile it, and upload it to my Mk3. Unless it is no longer open source, the calibration algorithm should be visible to anyone who wants to look it up.
1 Like
If your parts come out 0.18mm undersized no matter what size the total is You have very very severe underextrusion… That would make itself visible in other parts of the print too… (at 0.40 extrusion width 0.18 is almost half of it…) So I don’t think so…
Or do You mean 0.18mm on a 20x20 calibration block? If so that value (shrinkage) would be quite normal for ABS or ASA. You can fix that in OrcaSlicer
I always achieve accuracy below 0.03mm. But I’m still using the Prusa MKS3+ as I’m still waiting for Bambulab support’s answer to the non-functionality of my new X1C…
My basic rules for accurate parts:
- perfect geometry of the machine (correct assembly)
- frame reinforcements
- clean and flat surface
- dry filament
- calibration of extruder steps
- flow calibration
- slicing with option ‘outer perimeters first’
- high slicing accuracy
- no dimensional compensations
- correctly drawn object
I print with a 0.6mm nozzle, I reach a speed of almost 50g/h for larger parts. I am fully satisfied with the Prusa, but I wanted a compact design and a fully functional fine-tuned printer. I want to find out how good the X1C is, but without Bambulab support, I’ve only had a “nice pizza heater” for 1600€ at home for a week 
3 Likes
I just printed the CaliLantern from Vector 3D because I was noticing some slight skewing for some really big models I’m printing. I never saw it until I printed this large and had a join between to angular pieces that were showing a significant gap. I took all my measurements using a Mitutoyo caliper and using the spreadsheet that comes with the CaliLatern. I didn’t find HUGE imprecise numbers, but they were large enough that over enough distance (such as my prints) you would start to see issues. I can state very matter-of-factly that there is an skew issue, that I really have no idea how to fix since everything is closed source I can’t modify the FW to compensate.
1 Like
There is skew correction compensation being discussed in this thread: GCODE for Skew and Scale correction? - Bambu Lab X1 Series - Bambu Lab Community Forum
Maybe that will lead you in the right direction?
Thank you. I started that thread. I had an error when I first tried the Marlin skew correction, but turns out it was unrelated to the gcode change. I need to test again.
ABS for me shrinks in all axis, maybe this is why bambu has not added softfever’s shrinkage yet. I found a 70mm tall object comes out around 0.5mm shorter, so I scale the whole model. I’ve also had PETG shrink 1mm on a 180mm long part on my other printers.
Thank you! Yes, that is super helpful. Thank you for pointing that out.
My skew measurements came out similar to yours:
The one person who had replied to that thread had given me reasons as to why I couldn’t hope for better, and to leave it alone, but now, after reading some of the comments here, I’m starting to think that improvement might be possible after all.
All of which raises a worthwhile question: just how good can we realistically aim for, and at what point can we do no better and should just leave it alone?
At least to me, the CaliLantern required a fairly large number of tedious measurements. It’s not something I would want to do very often.
1 Like
That’s an excellent question and my off-the-cuff answer is: It should be as good as, if not better, than any open source printer. We are paying a premium for the BL printers and since they are closed, they are basically saying “we know what we are doing, trust us.” The stuff I’m printing is being printed by several other people on everything from Enders to Prusas, and they are getting tighter tolerances and results than what I’m getting. So, BL needs to either let us tweak to the FW as needed, or perform better quality control at the outset to ensure 0° skew since we can’t make the adjustments that need to be made.
4 Likes
I like your answer. The BL prints a lot faster than the Prusa, though, so I wonder whether expecting the same or better level of accuracy at higher speed is fair or not. On the other hand, the BL can certainly print slower if you ask it to. I haven’t tried that. Does it make any difference? Anyone know? Or is this a foolish question?
No idea, but I feel like they should be able to leverage the LiDaR to true up the skew just like they do the flow. It could draw a large box around the perimeter of the parts and then scan the corners for 90°. No idea how to check the Z axis though.
3 Likes
I obtain prints with around 0.2mm of shrinking when I use a layer height of 0.2mm. If I use the 0.16 High Quality profile, I achieve around 0.05mm of dimensional accuracy. Very curious…
1 Like
Maybe that has to do with the height of the model? The slicer can only divide it into an whole number of layers. So with 0,2mm layers, one more layer might be just above the model’s height, so the print is almost a whole layer shorter, while with 0,16mm layers it might work out much better.
Resonance Compensation / Input Shaping is prone to alter dimensions (especially if it is not done perfectly but I assume Bambu has that handled). So yes tehoretically Printing slower would get you better dimensional accuracy.
1 Like
It only affects the XY dimensions. The Z dimension remains the same.
Sounds like we need a new theory. 
What Do you mean? What theory?
Has anyone found an improvement regarding this issue?
I tried everything that came to my mind, but I am out of ideas now.
The only thing that helped to improve accuracy a little was the “belt tightening” method.
Anyway an 100mm L-shaped object is still constant off -0,1mm on one axis and 0,2mm on the other.