First CNC'ed Part Made on My PM25 Mill

[bakrch]

It is most likely due to improperly set preload between the two ball nuts. Not sure what software you are running, but you can likely compensate for it until you are able to fix. I have about .002-.0025 on each axis, and even that is enough to give an undesirable performance.

 

[devils4ever]

Yeah, I tried to tighten the coupler to the threaded section on the ball screw, but I think I might have made it worse.

I can push and pull along the Y-axis and get the table to move 0.005". How do I set the preload on the ball nuts?

I'm running LinuxCNC. Is there a way to compensate for it here?

 

[bakrch]

Yeah, I tried to tighten the coupler to the threaded section on the ball screw, but I think I might have made it worse.

I can push and pull along the Y-axis and get the table to move 0.005". How do I set the preload on the ball nuts?

I'm running LinuxCNC. Is there a way to compensate for it here?

There is a spacer in between the ball nuts that is likely undersize. You need to remove that and replace with a stack of belleville springs. To disassemble you will need to make a tool to keep the balls from falling out. Assuming you have 1605 size, I can hook you up with those dimensions if you need them.

Here is what my X-axis looks like:

https://imgur.com/UOkA1gd

I have a ton of work right now though and haven't gotten around to fixing yet.

Not sure about Linux CNC, but as popular as it is ... can't imagine that it doesn't have backlash comp.

 

[devils4ever]

Yeah, that looks pretty tricky. A spacer has to go in between the ball nuts? You can't just spread them apart and re-tighten the screws?

I did find how to add backlash compensation to LinuxCNC. A quick test seems to confirm that it works.

I plan on milling a 3.000" circle soon, so I'll see how well it does.

 

[bakrch]

Yeah, that looks pretty tricky. A spacer has to go in between the ball nuts? You can't just spread them apart and re-tighten the screws?

I did find how to add backlash compensation to LinuxCNC. A quick test seems to confirm that it works.

I plan on milling a 3.000" circle soon, so I'll see how well it does.

There is already a spacer in there, but it was not sized with care. The spacer is a 2pc split bushing and the key goes from one ball nut, through the spacer groove and into the other ball nut (see my vid for reference). You could make your own spacer to replace if you have the means. Belleville springs are a much simpler solution and will force the ball nuts to remain apart (preload) without the task of making a precision part. The only drawback is the trial and error to get the correct force to resolve the issue all without going too tight, which would prematurely wear on the system.

 

[JimDawson]

The other place that backlash can come from is the ball screw end bearing. That needs to be preloaded also.

 

[devils4ever]

The other place that backlash can come from is the ball screw end bearing. That needs to be preloaded also.

I never heard of Belleville springs. I'll have to look into that.

I think I have the ball screw end bearing as tight as I can get it. I'm not 100% sure, but I loosened the set screw on the collar, locked the saddle, and snugged the collar against the bearing while tightening the set screw. I've done this 3 or 4 times now.

I'm guessing getting rid of the mechanical backlash is preferable to having the backlash compensated for in the LinuxCNC software?

 

[JimDawson]

I'm guessing getting rid of the mechanical backlash is preferable to having the backlash compensated for in the LinuxCNC software?

Much preferable, backlash comp in software is OK(ish), but there is no substitute for a mechanically tight machine.

 

[shooter123456]

Does anyone have suggestions on how to do this differently?

My preferred method is using a set of low profile clamps (Something like Mitee Bite talon clamps) that hold on to a small bit of stock, machine the top side on all 4 sides + the face, then flip it and deck off the gripped portion. It usually requires oversizing the initial stock somewhat, but it can be avoided if you are careful.

I think I have the ball screw end bearing as tight as I can get it. I'm not 100% sure, but I loosened the set screw on the collar, locked the saddle, and snugged the collar against the bearing while tightening the set screw. I've done this 3 or 4 times now.

What kind of bearings are you using? If they aren't set up to provide preload, it doesn't matter how much you tighten it, it won't get any better. That also isn't a great way to tighten the ballscrews into place since even with locked axes, the ballscrews may be able to move them. I also found that the standard lock nuts aren't very good for locking the ballscrews and staying locked. The set screw doesn't hold well and damages the threads.

I made some lock nuts that do a much better job and are easier to tighten.

 

[devils4ever]

I got the kit from Dave at ArizonaVideo, so I'm not 100% what bearings are used. You can see some pics here, but I'm not sure if you tell from these pics.

I might try the Belleville springs if I can determine the correct ones to use. I did add a backlash term to LinuxCNC, but I don't think it helped much. I'm still seeing a difference of about 0.007" and 0.008" in the diameter.