Quote Originally Posted by Valfar View Post
Now from my understanding that means stepper motors driven by belts are not accurate enough (what's the exact value of this? 10 thou? 100?), and the way to go would be either a worm gear or a gear box. But the problem with these is the backlash they introduce, and a good quality one is out of the question for a DIY budget. Is that correct?
One data point based on my own recent experience. My setup might or might be applicable to your situation but it might give a little bit of insight. I've been working on a Z height setting macro on my router. I'm winding a 2.2KW watercooled spindle up and down using a 1605 ballscrew driven by a 3Nm stepper via a 1-1 HTD5 belt and pulleys. Max speed 2500mm/min but can't remember current acceleration settings off-hand - think it's about 400mm/s/s. I can touch off the setting plate, wind the Z up and down 50 times at full acceleration/rapid speed, and retouch. I get repeatability within one or two microsteps, time after time. That corresponds to 3-6 microns (0.003-0.006mm) variation. Ok, the machine is very new so the belt has hardly had time to stretch, and I don't claim that the machine has long-term accuracy of anything like this. Similarly, I can set one tool to the work surface, change tools, retouch off a fixed setting plate, and the new tool will be just touching the surface of the work. That is testing various aspects of the drive system, and I'm pretty happy with it. You will be needing very high precision gearing to get anything close to that, and/or very careful attention to fitting and adjusting and maintaining, and/or a lot of money to buy the components. Although my own experience is with a router, I would be happy to get that performance out of a vertical mill.

Belt drive might not be perfect but it beats any gearing available at an affordable price, I suspect. And while a stepper might not have the resolution of a high-count servo encoder, how often will you need that level of resolution?