What's new
What's new

Cons of rolled ballscrew vs ground for low accuracy cnc axis?

gregormarwick

Diamond
Joined
Feb 7, 2007
Location
Aberdeen, UK
We have a big old manual lathe with a 15" hollow spindle in decent mechanical condition that I'm considering doing a retro on. We have had to decline a fair amount of work recently that we could have done if this lathe was cnc. The retro itself doesn't phase me, btdt. And I have almost everything I need to do it on hand. I have a decent selection of high quality ground ballscrews that I can use for the cross slide, but I would have to buy one for the z axis. The lathe already has a scale that I could integrate easily enough so the accuracy of the ballscrew wouldn't be a big concern. A small amount of backlash probably wouldn't be a big deal either. Given the size of the machine, I'd likely need at least a 50mm fine pitch screw at 3m long, and these get pricey. C5 rolled is much more affordable, but I've never worked with rolled ones before. Besides the loss of accuracy are there any other major downsides to a rolled ballscrew that I need to consider?
 
Ground ballscrews have the advantage of stiffness (more uniform pitch, and consistent ball race = better load transfer among all the elements = high stiffness). Depending on preload, they can also be easier and more consistent during rotation, as variations in the lead and ball preload are smaller, so no "pulsing" or other form-induced variations.

Downside, of course, is cost. And at the size of screw you need, it's a big deal.

One thing to consider is that the machine could be designed to have a shorter ballscrew for carriage drive that can be moved and clamped to the bed, such that you have a "section" of CNC, then shift, re-calibrate, and now cut in the new zone. Much cheaper to implement (1 meter ballscrew is about a tenth or less a 3 meter), and with many features only requiring a relatively short CNC path, this can work in some applications.

Obvious drawbacks are not having full access to the entire bed for CNC control (but the standard leadscrew can be maintained for general turning), and you must reference features properly to maintain end-to-end accuracy. But for a big old beast like you're discussing, this may be a viable solution.

While I'm not an expert, I am reasonably knowledgeable about ballscrews, and have a ton(ne) of them here. If I can be of help, let me know. [but I'll be off the forum for a while today, so be patient]
 
Ground ballscrews have the advantage of stiffness (more uniform pitch, and consistent ball race = better load transfer among all the elements = high stiffness). Depending on preload, they can also be easier and more consistent during rotation, as variations in the lead and ball preload are smaller, so no "pulsing" or other form-induced variations.
Ok, thanks. So do you think that such concerns actually preclude the use of a rolled screw for my application?
One thing to consider is that the machine could be designed to have a shorter ballscrew for carriage drive that can be moved and clamped to the bed, such that you have a "section" of CNC, then shift, re-calibrate, and now cut in the new zone. Much cheaper to implement (1 meter ballscrew is about a tenth or less a 3 meter), and with many features only requiring a relatively short CNC path, this can work in some applications.
That's an option that I have considered, but really a last resort. I'd obviously prefer to keep the retro mechanically simple, and that setup would hamper the utility of the whole thing to a degree. I have a 1.5m screw on hand that would work well for that, but again, not really the way I want to go.
 
Make it take a std cheap but good properly preloaded nut ball-screw and let it make enough - earn its keep. If it becomes a pain, by that point it should have paid for its self well enough to upgrade any rate. My experience of rolled ball-screws is pretty good. Yes they have lead errors, but its not like they don't have good steady preload over there lengths, the pitch in my experiance of them remains pretty dang constant it just might well not be a perfect 5-10-20mm pitch. if you want good accuracy over 3 meters ground screws are going to need some error compensation any-rate. These lengths thermal effects become very very noticeable. If your planning to drive the screw with a servo and encoder then either software compensate or use the glass scale for on the fly compensation then i don't think you will find it much of a issue to save the money and go rolled. You don't mention lathe HP, but a 50mm screw sounds small for a machine of thoes sizes if you want to do things like carriage mounted drilling. A lot also depends on just what you want out of it, do you want a power chucked robot loaded production work horse, or just something more akin to a protract equipped setup that makes threading and more shall we say artistic - radius like shapes easy.
 
What the previous poster said was just what I was typing!

A ground ballscrew might be more expensive than a rolled screw with glass scale, etc. At those lengths you really have to pay attention to the pitch of the screw - if you get too tight a pitch you have to turn at at a high RPM and it will whip / deflect quite a bit. If you look at long ballscrews on machines their pitch is often equal or greater to their diameter (i.e. 40x40mm).

What type of controller are you using?
 
I've used both, and I would probably use a rolled screw if I were you. Good rolled screws are pretty darn accurate. One of the common inaccuracies is the pitch will be consistently above or below nominal; this is easy to comp out. Having secondary feedback really eliminates many/most of the issues of rolled ballscrews.

It's not applicable in this case, but I'll throw it out there for others, if you have an axis driven by two rolled ballscrews you can request that both screws have the same average pitch (at least you can from Bosch). Then you won't have on + and one - screw.
 
Make it take a std cheap but good properly preloaded nut ball-screw and let it make enough - earn its keep. If it becomes a pain, by that point it should have paid for its self well enough to upgrade any rate. My experience of rolled ball-screws is pretty good. Yes they have lead errors, but its not like they don't have good steady preload over there lengths, the pitch in my experiance of them remains pretty dang constant it just might well not be a perfect 5-10-20mm pitch. if you want good accuracy over 3 meters ground screws are going to need some error compensation any-rate. These lengths thermal effects become very very noticeable. If your planning to drive the screw with a servo and encoder then either software compensate or use the glass scale for on the fly compensation then i don't think you will find it much of a issue to save the money and go rolled. You don't mention lathe HP, but a 50mm screw sounds small for a machine of thoes sizes if you want to do things like carriage mounted drilling. A lot also depends on just what you want out of it, do you want a power chucked robot loaded production work horse, or just something more akin to a protract equipped setup that makes threading and more shall we say artistic - radius like shapes easy.
Wouldn't expect to be doing much drilling with it, it has a geared tailstock if I need to do that, also it won't be in 24-7 production, occasional jobs only. It is special only in it's big hollow spindle, it's not a particularly huge lathe otherwise, 2-1/2' chuck, 22" over the cross slide. I'd likely end up doing some semi-heavy roughing on occasion, but nothing too brutal. Think 10mm DOC, .8mm/r in alloy steels, that would be fairly typical of what we do on it right now, a small downgrade in that dept wouldn't be a big deal tbh, I need it to have contouring ability, not too worried if it means small cuts. I'd be using a DC servo with tach, and using the linear scale directly to close the loop. Rotary encoder on the cross slide tho since I have "real" ballscrews I can employ there.
 
Ok, thanks. So do you think that such concerns actually preclude the use of a rolled screw for my application?

Not preclude, just factor in if that's the way you want to go. For a long screw make sure you're got concentricity/straightness values that don't help lead to whipping.

That's an option that I have considered, but really a last resort. I'd obviously prefer to keep the retro mechanically simple, and that setup would hamper the utility of the whole thing to a degree. I have a 1.5m screw on hand that would work well for that, but again, not really the way I want to go.

With a longer screw you have to factor in critical speed and thermal correction, and it's also helpful to have a stretched mounting, which is not practical with a moving mount. If you do consider a sectional CNC, I'd keep it to a meter length.

What else you should make sure of is carriage and cross slide movement and lubrication, including whether you want Bijur style pulsed lube or continuous. As a plain-way machine, you may want to confirm no major wear in the bed near the headstock. If the cross slide has worn so that you have loose areas when the gib is adjusted to the least worn section, this can be an issue too. Wipers, shielding the ballscrew and nut from chips matters too.

Post some pictures of the machine, with details of the carriage and bed, along with what your preferred layout is. Some feedback here could help.
 
According to the research I have done recently in my own search for a ball screw and nut, modern rolled screws are almost as accurate as ground ones, and if I remember correctly, because of the forces imparted on them during the rolling process they are usually more stiff than a ground unit. Lead accuracy is very close to a ground screw and with all the ball nut variations available, zero backlash is pretty easy to achieve.

Stuart
 
What the previous poster said was just what I was typing!

A ground ballscrew might be more expensive than a rolled screw with glass scale, etc. At those lengths you really have to pay attention to the pitch of the screw - if you get too tight a pitch you have to turn at at a high RPM and it will whip / deflect quite a bit. If you look at long ballscrews on machines their pitch is often equal or greater to their diameter (i.e. 40x40mm).

What type of controller are you using?

I was considering a 50x10. I don't want to run out of torque obviously. I don't care about rapid speed and the spindle maxes out at 300rpm so there will be no high feedrates involved either. I haven't spent any time working any of this out yet but I just can't see it ever spinning very fast.

I would use LinuxCNC with mesa hardware. I have a full functional system on the shelf that I removed from a previous retro after it had served it's purpose. I have a pretty good selection of big old dc servos and amps to choose from.

One of the common inaccuracies is the pitch will be consistently above or below nominal; this is easy to comp out.

Good to know, thanks. If that's true then it will indeed be easy to comp for.

Not preclude, just factor in if that's the way you want to go. For a long screw make sure you're got concentricity/straightness values that don't help lead to whipping.



With a longer screw you have to factor in critical speed and thermal correction, and it's also helpful to have a stretched mounting, which is not practical with a moving mount. If you do consider a sectional CNC, I'd keep it to a meter length.

What else you should make sure of is carriage and cross slide movement and lubrication, including whether you want Bijur style pulsed lube or continuous. As a plain-way machine, you may want to confirm no major wear in the bed near the headstock. If the cross slide has worn so that you have loose areas when the gib is adjusted to the least worn section, this can be an issue too. Wipers, shielding the ballscrew and nut from chips matters too.

Post some pictures of the machine, with details of the carriage and bed, along with what your preferred layout is. Some feedback here could help.

I doubt that I will be able to stretch in any case. The extant mounting points on the bed casting are too far apart, to use them would mean a much longer screw than is actually required. As such it will most likely be anchored at one end only and some kind of self aligning plain bearing support at the other. Thermal growth of the screw should be a non-issue if I go with direct linear feedback I'd have thought? Deflection of the screw under thrust is something I am concerned about though. It is only 2.5m of actual travel so I'm thinking the 50mm screw should cope? As far as lube I have a nice vogel gear pump unit that I would use on a timer. Slides are in good condition, otherwise I would not even entertain this idea. They have been reground and refitted with turcite/moglice at some point (before I arrived here) and not used a whole lot since.

According to the research I have done recently in my own search for a ball screw and nut, modern rolled screws are almost as accurate as ground ones, and if I remember correctly, because of the forces imparted on them during the rolling process they are usually more stiff than a ground unit. Lead accuracy is very close to a ground screw and with all the ball nut variations available, zero backlash is pretty easy to achieve.

Stuart

The lead accuracy on the ones I am looking it is nothing to write home about. Seems that ≥50 dia most places only offer 50µm/300mm. Again, shouldn't really matter with direct feedback, but that's a big departure from a good ground screw. If the error is consistent as DanielG posted above then it's simple to comp for, but if it actually does vary the +/- 50µm over the length of it, not so easy.
 
Last edited:
It is only 2.5m of actual travel so I'm thinking the 50mm screw should cope?

Think of the ballscrew as a long flexible thing. It's really critical where you put the fixed end and how it's supported. If you can't tension it you loose of stiffness. One option would be to try and position the fixed end / mount near the headstock- this way you are pulling on the ballnut and as you take a cut towards the head.

If you mount the fixed end mount near the tail end of the lathe, you are always "pushing" on the ballscrew when you take a cut towards the tailstock. Any idea what the weight of the moving parts and cutting forces are? That will help a lot in determining the ballscrew size.
 








 
Back
Top