Acceptable backlash in C-Beam machine?

Hi guys, greetings from Italy.

I've recently built the C-Beam machine, and after some unsatisfactory runs I've decided to tear it apart and reassemble it more carefully. I have zero experience in machining, and my knowledge in the CNC world is limited to 3D printers (which are far less demanding from a mechanical point of view) and a very small router for milling PCBs.

My question is: how much backlash should I expect from a well-tuned machine, especially in the X and Y axes? Despite my best efforts, I can't get backlash to go below 0.05 mm, and this is with the anti-backlash nuts installed. I'm checking this with a (seemingly good-quality) dial indicator with 0.01 mm resolution.

Local accuracy of linear motion is consistently within +/- 0.015 mm when moving in one direction, so the 0.05 mm backlash when inverting the direction is disappointing... Is there anything I can do to improve the backlash figure? Further turning the screw on the anti-backlash nuts doesn't help (actually, it makes things worse from some point on).

Cheers,
Marco

 

Thanks for the advice, dddman. I've tightened the lock collar a little bit more, now backlash is around 0.03 mm. Further tightening doesn't eliminate the remaining play, it just makes the lead screw wobble. There must be another source of backlash that I'm missing...

 

@dddman: end plates are tight, and everything seems to be stiff enough (I can't feel any loose parts). I've spent a fair amount of time trying to get the best configuration for eccentric nuts on the wheels so that there is no play. Initially I had tightened the nuts too much, and motion was irregular at the 0.1 mm scale, so I gradually loosened them until movement became precise again. But still, that small backlash was there. According to the manufacturer's website, the dial should have an accuracy of 0.007 mm, and repeatibility of 0.005 mm. For X and Y measurements I mount the indicator on a 90° bracket which is then secured to the v-slot, as in the picture below:



As far as I can tell, the dial indicator is parallel to the beam.

@puntoMX: I'd be more than happy to get an overall accuracy of +/- 0.1 mm on machined parts, but the backlash issue is just one source of error. When milling metal, for example, the whole structure is likely to flex quite a bit. With the dial indicator set up as in the picture above, if I try to manually pull the spindle shaft in the +X or -X direction with a force of 20N (measured with a dynamometer), I get a displacement of +/- 0.05 mm. So, if we take into account the not-so-rigid structure of the C-Beam machine, the accuracy and backlash of the whole drive chain, vibrations and resonance, and other sources of error that I probably don't even know, the overall accuracy is likely to be worse than 0.1 mm. And this is without even considering the absolute error caused by non-perpendicular axes, which is the second issue I'm trying to solve on my machine.

 

Radial play seems to either be very low or not affecting the measurement anyway, because when I manually pull the lead screw in its radial direction I can barely see the indicator's needle move.

Anyway, I think the issue lies somewhere else. I've just tried to rotate the motor shaft manually, and guess what? Backlash is less than 0.01 mm. I've repeated the test several times (it's easy to feel when the shaft "snaps" at fixed one-step intervals), and that value is consistent. So the larger backlash figure is probably due to the motor itself, or maybe the driver...

 

@dddman: nope it doesn't. I've tried speeds from 10 to 3000 mm/min, and accelerations from 1 to 1000 mm/s^2, and I always get the same values.

 

Yes, same amount of backlash on all axes. Precision (not only related to backlash) is slightly worse on the X axis, probably because the X lead screw wobbles more than the other two (it came already warped).

I must say I don't trust my motor drivers, as they have already shown to be far from perfect (cheap chinese stuff based on TB6560). I can't run them at the maximum current allowed by the motors because motion becames very irregular, and the whole frame resonates badly. So I'm actually driving the motors underpowered, which might also introduce additional backlash due to loss of torque between microsteps (at least that's what I'm reading on the web about the topic). I'll try to drive one of the motors with a good old DRV8825 and see if anything changes.