Gantry wobble - which screws do I have to tighten?


I finished my MPCNC and first tests look very promising:

Today I wanted to try some milling and I saw that the spindle is “wobbling” a bit. It seems my gantry has some “play”. Please have a look at the video (it is hard to show what I mean). It is a very small movement but I think it is bad when milling. My question is, what screws do I have to tighten to make the wobble go away or less? Thank you!!


1 Like

Hard to say. You bearings are not touching. They all should on there own and when you assemble it it should get tighter. As much as everyone hates to hear this it almost always works perfect the second time. Take it apart swap a part and reassemble it and pay close attention as you go. I bet it will be near perfect the next time…or you have bad prints.


Ryan, I am at almost the same point with my renewed machine. All bearings are in touch, the whole construction seems to be much more ‘on the point’ than the very first one, see elsewhere.

Today, I measured the angle of the X-Y-gantry rails, first completely unloaded on a desk, then put loosely on the rollers. Same result, the angle (measured in the quadrant where the tubes meet directly, see from above) was 90.7 degrees. Controlled playing with bolts A and B led to results between 91.0 and 90.5 degrees. Whatever combination I tried, I could not get below that value. The two A boltes are so loose now I can turn them by hand. Snugging B to the limit (no cracking sound yet!) did not really change much.

As I use Linuxcnc to control the machine, there is a theoretical solution, but my Linux- knowhow reaches its limit on this too quickly to let it look like a practical workaround:

Shall I go the empirical instead of the deterministic way and disassemble everything again?

The parts were printed on an alfawise U30 which I triple-checked for squareness. PETG.

BTW, the 130mm bolt can easily be replaced by a 120mm one. Less protrusion, more pleasant to the eye.

Did you have the Z axis in place, it should be to check? 10 degrees off is wayyyy to much without something being wrong.

Ryan, assuming you meant my question: The error is 0.7 degrees, downsqueezable t0 0.5 degrees, but still too much in my eyes. Yes, all tubes in place, fully inserted.

Sorry, that was pre coffee. So I would say loosen the 1.75" and the 5", make sure the 2.5" are just starting to touch and slightly snug (holding the Z axis). Run it a bit and see which way it settles. If it gets worse and can’t be adjusted, for whatever reason a part swap(switch the XYZ parts top to bottom) and rebuild almost always fixes it.

The dual endstops can easily handle under one degree, but a accurate build is always the better choice.

1 Like

The 40mm screws are “snug” here so far, so you s(n)uggest to release them too a bit? Geometrywise, that makes sense. I’ll try that. In the end, it is not that much of way close to the center, but a few millimeters offset over the long axis. Sophisticated mechanics… :slight_smile:

It is hard to describe but when I build one I end up using my arms as a stand in to figure out which tension bolts to mess with. Tightening them moves the rail away (and the reciprocal), keep the matched pairs under even tension.

Today, I found out about a possible reason for unexplainable angle errors, maybe. I measured the diameter of my Z-Axis’ tubes (polished stainless steel, specified to be 25mm diameter). Well, depending on the angle around the tube (measured at the same longitudinal position), the values ranged from 24.89 to 25.02 mm.

(I am sure I sent a similar reply earlier this evening, but it seems to have gone somewhere, but not here.)

Just spin them until you get a snug fit, Ovals do have an advantage like that!

Not quite trivial for the Z tubes, with the holes in place.

But spinning the cross tubes is an option. The values are all lower than 25.00 mm, and the tubes don‘t have an oval profile necessarily, rather ‚some shape‘.

The combination with the tightening and loosening of the bearings remains a key. As always, getting the last 1% of precision eats 99% of the time.
I’ll probably switch to the Rambo 1.4 solution.