Machining Ball Screws Accurately

How does one ensure minimum TIR on a rolled ( not purely ground..) ball screw before machining down for bearings, pulleys, etc?

Using a circular ( cylinder style) ball nut to clock on its outside definitely does not work!

Joe

could you turn a very close fitting sleeve to push over the screw and clock that true

Grease,free balls in balscrew and ER collets?

Just hold on the od of the ball screw to machine the end. If you use a bearing pack only on the drive end of the screw it will be adequately supported in most cases. It may be you have centres in the ends which you could use. Chasing microns is not necessary unless you can achieve the manufacturers machining tolerances, a runout of 0.005 is required on the ends with a suitable bearing shaft ground tolerance.

I recall Sir John saying somewhere that you stick 'em in a collet.

Mine were small enogh diameter to pass through the mandrel of the Myford. The end was held on a collet and I had turned down a hardwood shallow taper with a sliding clearence hole through the centre which pushed into the other end of the mandrel to centre the lead screw as it exited and act as a support.

regards Martin

It's a sort of chicken and egg thing - I cannot find specs , even on SKF rolled ballscrews, as to the outside diameter accuracy - its the ball bearing surface that is 'accurate', and not the rolled lips on the other diameter. So if that diameter varies by a 5-6 hundredths ( of a mm..) how do you grab that in a chuck ( of any sort) and clock the bearing surface? I guess a close fitting sleeve will be as close to clocked as the rolled lips will permit...

Its a 4mm pitch screw, 16mm OD, so I set it up in a set-true 3 jaw, with about 40mm protruding, and set up the DTI in the tool post, and ran the spindle real slow, maybe 3seconds/rev, with the lathe set for screw cutting a 4mm pitch thread. I set the DTI ball tip nicely on the ballscrew bearing surface. The concept was nice, but I spent hours chasing TIR - the results were repeatably inconsistent! - I could get it down half a division (0.005mm??) over 3 or 4 threads, but if I rotate the ballscrew 180deg in the chuck, its all out - 0.02 to 0.05mm. Rotating the screw back to the clocked position would show anywhere between 0.02 and 0.005...depending exactly how it nestled back into the chuck, etc.

Maybe I am being picky - just clock a good 16mm shaft up, set the chuck true, grip the ballscrew and machine it and be done?

Or use 'proper' fully ground ballscrews..

Joe

I think the point being made is that the thread crests cannot be used as a datum, because of the method of manufacture: rolling. Thus the thread's flanks or root should be used, hence Niel's suggestion of producing a datum 'outer surface' by filling the thread with balls. Not sure that I like the idea of balls in a multiply-split collet, though. Also, one would then be stuck with a non-adjustable runout-error due to the combination of inherent inaccuracies.

So, balls in the thread, sleeve outside and sleeve held in 4-jaw? But getting the sleeve ID correct means measuring over (3) balls in the thread, which is awkward. A quick-and-dirty method worth trying might be to use a DTI, with probe running in the thread, mounted on cross-slide, and thread pitch set up on change-wheels or gearbox.

[edit: was going to be the third post, but got comprehensively beaten to it! You guys are quick!]

Edited By Kiwi Bloke on 17/11/2020 20:18:53

Its something I have often wondered about even when machining normal threads.

The true centreline (axis) of a cut or rolled thread is not always concentric with the crests. I dont know whether that applies to ball screw threads but for best accuracy it would seem best to devise a way of using the ball track as the reference.

One way of getting threads concentric is to wind a suitable gauge of wire into the thread before putting it in the chuck so the jaws grip the OD of the wires which in turn are sitting on the thread flanks. Ballscrews would need mighty thick wire to do the same trick so,

What about having a thin sleeve long enough to contain a string of balls over several thread pitches and then pressing the screw with sleeve and balls fitted into a freshly bored hole in the lathe. Would need a dog to drive the screw.

Ian P

Slow typing on my part

Edited By Ian P on 17/11/2020 20:25:32

Maybe I haven't understood the 180 deg rotation of the ballscrew in the chuck, but isn't that what you'd expect, if the crests can't be relied on as a datum?

I think I'd rotate the chuck by hand, not run under power. Beware backlash (of course...). Since the probe's tip bears on an inclined surface, any axial error will appear as a radial error, perhaps multiplied. What do you know about the screw's and your lathe's leadscrew cyclic or overall pitch error, and is the lathe leadscrew oscillating axially as it turns, because of its thrust bearing error acting like a swash-plate?

Edited By Kiwi Bloke on 17/11/2020 20:44:27

When I worked for a jack manufacturing company we had a split sleeve with matching internal grooves that was screwed on to the ballscrew, pushing out a ball retaining sleeve. The whole setup was gripped in a collet with a supporting guide bush in the tail end lathe bore. A lot of work to make the split jig and all the bits but when you had about 100 to machine each day every little helped!

Wind some wire around the thread and and grip the wire in a collet. The wire diameter should be slightly greater than the thread depth.

Martin.

Is there any opportunity to develop this method and use soft jaws, each jaw counterbored for one or more balls of the correct diameter to suit the screw, with the counterbores spaced at 1/3, 1/4 or 1/6 of the thread pitch apart to suit number of jaws and 1 pitch apart if multiple balls per jaw?

Especially with a 4 jaw independent chuck, you would be gripping on the thread's pitch diameter and could then dial in concentricity in the normal manner.

Its the last sentence that is the problem, See Joe's post about clocking

Something like this:

Thanks for all the suggestions - Methinks this is going to be an exercise not to be thought about to deeply!

I am off fitting Lion collars again - back in 10 days so won't be able to login for a while.

Thanks again for all the ideas!

Joe

Thanks for that. I knew there had to be a flaw in something so simple.

Can I run another option past you? Select a 3-jaw self centring, set-tru chuck with the ball(s) in the jaws. Calculate the minimum gripping diameter of the balls when they are clamped on the pitch diameter of the ballscrew. Grip and dial in a circular rod of that minimum diameter. Remove rod and replace with ball screw.

A possibility if there is no set-tru available is to put a grub-screw behind each ball and use that for dialling in (slow the balls down with a bit of silicone sealant).

The logic is gradually taking us back-in-time

[observant readers might note my use of hyphenation for the purpose of disambiguation]

I suggest that we are on the verge of reinventing the “Box Chuck”

**LINK** https://www.model-engineer.co.uk/forums/postings.asp?th=106487

MichaelG.

.

Edit: just noticed that the price link is absent from that previous thread

... it’s of little relevance, but here’s one from Cousins:

https://www.cousinsuk.com/product/lathe-chuck-set

Edited By Michael Gilligan on 18/11/2020 10:18:12

I thought the problem was being able to clock it true not actually holding the part. How would you clock it once in the box chuck Michael?

Joe, you may be better off buying screws with a reduced plain spigot and/or ctr drilled hole already at the ends and clock that assuming the makers would have made those ref points true to the ball grooves.

Edited By JasonB on 18/11/2020 10:35:40

As we're all supposed to be engineers here, why not ask the question, "Does it matter?" Bit heretical, I know, and I like fussing about details as much as the next person, but what is the real issue with a little bit of eccentricity? First of all, we're dealing with rolled, not ground, threads so these aren't the most accurate ballscrews in the world. Perfectly adequate for this job, maybe - so does a bit of out-of-roundness significantly change the overall system accuracy?

Secondly, it might matter what is going on the turned portion. Bearing plus drive pulley? How close does the ballnut get to the bearing? What is the radial tolerance of a rolled thread anyway? Those two affect the bearing more than the pulley. What longitudinal error is created by a slightly eccentric drive pulley?

My approach would be - don't want to use a chuck, even a Griptru or similar, as I don't like hardened jaws on a hardened workpiece. Grip can be unreliable. Collet is highly preferable. So it's a 5C in the lathe, packing at the back end of the spindle to stop it rattling, and off you go. Actually, my worry is the next bit - can you machine the ballscrew through the hard outer layer and get deep enough for a good machined surface in the softer metal underneath, including the bearing clamp nut thread? I've seen plenty of opinions on the best way to do that but haven't done it for myself yet...