I have a strange problem with my machine which is essentially a large Ultimaker using larger linear motion components as well. I printed these two cubes as a test to get a better understanding of what is going on.
One cube is 20mm and the other is 40mm. The two were printed at about the same Y coordinate but spaced apart in X. The 20 mm cube came out about 0.2mm undersize in X and 0.4 mm undersize in Y. The lower part of the 40 mm cube came out at 39.9 in X and 39.5 in Y. The upper part is noticeably undersized but has also drifted to the left some as well.
Other noticeable defects are disconnected top infill on the 20mm as well as a noticeable offset at the start/stop corner of the 40mm. Where the large shift occurs in the 40mm, the infill is also disconnected from the right side wall. Circles do not come out circular either.
I thought this problem was related to mechanical lash so I went from 6mm GT2 to 9mm GT2 on all four belts. I also went from fixed pulleys at all shaft points to idlers where I could. Each belt rides on only one fixed pulley. Since only two of the outer shafts have to rotate they are the only two using plain bushings and the other 4 shafts use linear ball bearings. None of these improvements made a noticeable difference in the dimensional accuracy. The machine uses Nema 23 steppers at 2.4 amps. I have never heard the machine skip steps nor does this look like such a problem to me. Using a Duet WiFi controller.
I would suspect a loose pulley. Not very loose, but just loose enough to cause a periodic problem. Depending on just how many drive pulleys you’ve got it could be a few different spots, too, but that would be my guess.
I think ultimaker style is very sensitive to alignment of the bearings/bushings. Try disconnecting motors/belts and running the gantry by hand. Make sure it feels smooth and tight for the entire X/Y space.
@Stephen_Baird The problem isn’t really periodic. In this case I can see where it might look like it is from this experiment. All parts I print individually come out undersized by about a half to full millimeter in X and Y depending on the size. In this case, it seems that having the two parts printed at the same time and separated along X reduced the error of both of their X dimensions but the Y dimensions of both are still undersized. Lastly, you see that the X dimension of the 40mm cube both got smaller and the layers shifted to the left once the 20mm cube was completed.
@Oystein_Krog If there are no belts or motors attached everything runs very freely, granted so long as the plain bushings are not bound by pushing at one end. To reduce the possibility of this I went to linear ball bearings wherever possible. I’ve made fixtures that put the outer shafts at the proper vertical spacing. The cross shafts are only fixed at end too. The other end floats in a closely fitting hole so that misalignment of the outer shafts of an axis will not cause binding.
If the undersize error is consistent across filaments and is steady between prints I would suspect it’s a calibration error in steps, although a very small one. That may be down to variance in the pulleys from expected size and would be best dealt with my calculating the correct steps based on your measurements, but only if it really is consistent across all prints and it’s a material specific shrinkage rate or something.
The shift still seems suspiciously like a loose pulley, though. There may have been a fast move (and the only fast move in the whole print) after the completion of the shorter cube which gave the printhead enough momentum to shift that loose pulley just a bit.
Given the size of steppers you’re using this is even more possible because they can exert a great amount of stopping force, making any slightly loose pulley more likely to shift where it wouldn’t be able to on a lower powered machine that would either have to move more slowly or would absorb that momentum in positioning error.
@Stephen_Baird I have tried using PET and get the same problems. I’ll upload a picture of results a part size vs error experiment I did.
@Stephen_Baird
missing/deleted image from Google+
Hm, so the problem is persistent, but the error isn’t consistent.
@Stephen_Baird Yes. As a side note, that hopefully doesn’t confuse the problem any more, I have tried bypassing the onboard X Y drivers with two Gecko stepper drivers. But with those, the machine ends up building parts at about a 30 degree angle from vertical in both X and Y. I don’t know if that issue is related to this one or not but I never resolved that either.
That’s probably unrelated, I know Gecko drivers can be temperamental about the pulse length for their steps so you may be experiencing missed steps when you use them as external drivers.
