We’ve got mixed feelings about a new video from [AndysMachines] that details how he makes custom ball screws. On the one hand, there’s almost zero chance that we’ll ever have an opportunity to put this information to practical use. But on the other hand, the video gives a fantastic look at the inner workings and design considerations for ball screws, which is worth the price of admission alone
The story behind these ball screws is that [Andy] is apparently in cahoots with SkyNet and is building a T-800 Terminator of his own. Whatever, we don’t judge, but the build requires a short-throw linear drive mechanism that can be back-driven, specs that argue for a ball screw. [Andy] goes through the challenges of building such a thing, which mainly involve creating threads with a deep profile and wide pitch. The screw itself wasn’t too hard to cut, although there were some interesting practical details in the thread profile that we’d never heard of before.
The mating nut was another. Rather than try to cut deep internal threads, [Andy] took a sort of “open-face sandwich” approach, creating half-nuts in a single piece of brass using a CNC machine and a ball-nose mill. The threads were completed by cutting the two halves apart and bolting them together — very clever! [Andy] also showed how the balls recirculate in the nut through channels cut into one of the half-nuts.
Whether the results were worth the effort is up to [Andy], but we were just glad to be along for the ride. And if you want a little more detail on lead screws and ball screws, we’ve got just the article for that.
I’m definitely going to follow this. Ball screws are crazy expensive. A bit cheaper if ordered straight from China, but have heard too many horror stories about people receiving them bent.
looks like he mainly focuses on the nut, which is the harder and more important thing to do, you can have a bent screw as long as its not wildly out of tolerance can give the nut a fighting chance
but lol yea unless you have some god like powers of the lathe machining any screw at a significant length is going to be bent or tapered somewhere
Case in point, I have one of those tiny tiny toy lathes you can get on ebay and whatnot for about a 100 bucks, and its cross slide screw gets a bit too tight in the middle no matter how I adjust its (lol) nylon gib. No worries mate we have a extremely nice PM, not mini, not engine metal working lathe at work, Ill just turn up a new screw
its 10x worse, but somehow measures quite a bit better over its little itty bitty 3 inch ish length, whats the problem OH the thread it goes into is off kilter! screw it im just using it to cut off plastic, brass and nylon PCB standoff’s and mating screws
Did you use a travelling steady when you were screw cutting the thread?
Well I never heard of a ballscrew before, it’s nice to learn something new. I also never knew hardware could sound so dirty, also fun to learn.
Ballscrews are the motion control components for 95% of all CNC machinery on earth. They are super common but very expensive to make properly accurate. The ball screws in most serious CNC machines are precision ground for the thread profile, with slightly preloaded oversized balls of a very high accuracy used to fill the ball nut.
Cheaper ball screws are created using precision thread rolling, and are usually cheaper to make but you lose some accuracy positional.
These are extremely common items in industry but they are wear items and they wear out. I think this might be the first time I’ve actually seen somebody make their own though!
Very nice. That back-driving demonstration at the end was quite impressive.
Also:
“building a T-800 Terminator of his own. Whatever, we don’t judge”
I judge, that’s cool as hell.
Ah yes the classic “I’m totally not building a lady-bot in the basement, it is just a Terminator” ruse. He isn’t fooling anyone- just own it man.
A SFU1204 costs around EUR 15 from China, and for that you have a hardened steel spindle + nut + balls + wipers. Attempting to make them yourself may be just for the fun of it, but it’s not efficient or cost effective. Brass is a terrible material for a ballscrew nut. For ball bearings used on any significant load you really need a hard material such as hardened steel. Soft materials WILL get deformed by the small contact area of the balls.
Those rolled ballscrews are quite affordable. Ballscrews start becoming expensive only when you start needing very tight tolerances and precision.
Are you German by any chance?
Try watching the video before you comment
I did see most of it, but was annoyed by the constant silly jokes. A few jokes is OK, but it got boring after a while. In itself it’s sort of impressive he made his own, and creative of how he did it, but (at the end) he also confirms that he can’t make it even close to the quality of a bought one.
He also stated he could not find one that fits his needs. A small search shows you can go down to SFK0601 (6mm diameter 1mm pitch) on Aliexpress. for around EUR 60. But they do tend to get a bit more expensive when you go below 1204. If you need a really small nut, Some models don’t have the big flange, and you could even grind it off.
So, a hack and worth of a Hackaday article: Yes.
Good Idea to reproduce: No, unless 2 days of machining to make something of a lower quality then you can buy for EUR 60 is your idea of a hobby, or if you’re a youtuber and you do it for the clicks.
And I do hope the ball screws he made are fit for his purpose, but I also won’t be surprised if he gets back here next year with a “fail of the week”, because his home made ballscrews are not good enough for his own application.
The beads are hard steel and the nut is soft brass. How can’t the nut be eaten away? How can the device keep it’s precision?
Try watching the video where that’s explained.
Welcome to HaD. you must be new. Lemme get you up to speed with other classic comments you can feel free to use:
1. Could have just used a 555
2. Not a hack
3. Title is clickbait. AND not a hack, also could have just used 555
.
feel free to riff on it but you get the idea.
cheers
this comment is both very sad AND very true
You forgot “this is hackaday, not buy-off-of-amazon-aday :D
You do realise that is an inevitable ‘failure method’ somewhere along the line as zero wear is impossible and one part will always either be taking loads in a worse way or be the more fragile material. The screw is going to wear too…
In this case the Brass is going to last pretty well I’d suggest – yes its softer but will polish with a bit of wearing in to a very smooth, well sized and work hardened finish that suits the balls and with all the lubricant the two metals are most likely not actually in contact with each other much. There is after all a reason we use lubrication!
FTR, at first brass will give slightly against the balls, but eventually the brass will last longer than steel bearings in a wear environment, and this is always true, because of lapping action. As even the smallest amount of grit inevitably gets into that system, it will embed in the softer material of the 2, which then long term becomes a lapping object. As its softer, its compliant more than it wears, and will hold the harder grit and wear down the steel balls.
This is true in watch and clockmaking- even on timepieces several hundred years old, when looking at which is worn, the larger brass wheel gears or the smaller steel pinions, the pinion gears will always be where the most wear happens, for the exact same reasons.
It’s the opposite of what most people think should happen based on hardness, but its a wear mechanism and failure mechanism that’s present in all mechanical systems with materials of different hardnesses. Look into the study of wear- Tribology. You’ll see this is true.
Very nice work, but for a backdrivable way to get linear motion from rotary motion, why didn’t he make himself a rack and pinion drive? I’m designing a CNC machine right now, and am going to have ballscrews on it (<£40 for fairly short 1605 sized kits, I’ve been testing one and you can backdrive it but it resists that quite hard) because they’re the “precision” option, but for a robot which doesn’t need that much precision but where backdriving does matter a rack and pinion makes more sense. For smoothness he could always herringbone the gears, and he can use reduction gear stages before the rack and pinion if a typical gear ratio for a rack-and-pinion doesn’t give as much linear ouput force as he wants per unit of motor torque.
Because it would be harder to make it look like hydraulic actuator that he needs for his terminator robot.
There was a CNC lathe in our shop that had rack and pinion drives. It was used for multi pass threading on screws with pitched up to 10 inches.
The drives had two pinions preloaded against each other to eliminate backlash.
In spite of the severe service the machines ran for decades without need to replace the drives.
Ballscrews on machines seeing similar service would last about three years.
Cool plaything, but unfortunately not usable for anything that requires precision.
Firstly, the title is misleading. These ball screws aren’t rolled, they are cut (on a lathe here). That means that the tolerance of the ball screw is entirely dependent on the tolerances of the lathe and the cutting forces involved. Tolerances within the gibs and ways, or linear rail systems, will affect cutting head positioning. Tolerances within the lathe screw and gearing, or ball screw and motor, will also affect cutting head positioning. Cutting forces will cause deflection, although with good machining practices that can be managed. To cut a long story short, it is impossible to end up with a cut ball screw that is more precise than the lathe that it was cut on, and the precision will change along the length of the ball screw.
With a commercial rolled ball screw, the screw is formed by rolling the stock rod under high pressure between two tapered roller forms. The screw precision is determined by those forms. The key is that the feed rate of the stock is effectively determined by the the thread pitch on the forms and their rotating speed, and that all points on any part of the circumference of the screw get formed by the same points of the forms. That means that the precision along the screw cannot vary much and is nearly entirely determined by the quality of the forms. They themselves are ground hardened material. Roll-formed ball screws are typically limited to P5 tolerance class, 23um/300mm, although the better quality manufacturers can usually roll to P3 (12um/300mm) for a price. I got quoted over $2k for a P5 rolled 16mmx0.2″x19″ by Thomson, so I can only imagine what a P3 would have cost. Admittedly some of that was because they considered it a custom order, being a replacement for a ball screw originally made by a company they had absorbed that used rectangular nuts. Fortunately I found a replacement on ebay, and didn’t need to resort to making my own! The rolling process work-hardens the screw to some degree.
The traditional alternative to a rolled ball screw is a ground ball screw. A hardened stock rod is passed and rotated past a grinding wheel that has the desired thread form, with traveling steady rests positioned both sides. Cutting forces in grinding are very low, so deflection is minimized, and cutting typically only occurs in one direction, eliminating any backlash issues. Furthermore, grinding machines always typically have high precision motion components because they are dealing in tenths of thousandths of inches (0.0254mm). Grinding is very slow though, so ground ball screws are extremely expensive. It is still used for the highest precision ball screws.
In recent years a third method has become popular, splitting the quality difference between rolled and ground. Ironically it’s remarkably close to the method used in the hack – cutting. Whirled ball screws are cut on something that looks like a lathe. Yes, whirling has all the potential negatives that I mentioned in the first paragraph, but modern whirling machines work around them. The cutting head is circumferential, creating opposing cutting forces that reduces deflection. The cutting head rotates around a hardened stock rod as it moves along it, with steady rests on both sides keeping everything still. Cutting forces can be large, and cutting tools are typically carbide or harder (as would expected for hard-turning). Modern production machines often don’t rely on ball screws for positioning feedback but use a secondary system such as a linear scale, or laser interferometry. You can buy linear rails with built in linear scales with a 0.5um precision. Whirled ball screws can easily achieve P3 (12um/300mm) tolerances, and can be cut fairly quickly.
In all cases, commercial screws end up hardened. Induction hardening can be applied to rolled screws to ensure that they are harder than the balls. Grinding requires hard material to start, and whirling usually is hard-turning. That could be done on soft metal with post-whirling induction hardening but since the whirling machines are by design capable of hard-turning, it’s usually cheaper to buy hardened stock than post-harden in-house.
As for ball nuts, these are nearly universally cut and ground from a hardened steel. They are critical in achieving precision, and ideally need to be harder than the ball bearings (which are relatively cheap to replace). There are several methods of reducing or eliminating backlash. Ball sizing can be tweaked to adjust preload (at the cost of added friction). A traditional double nut pushes two different ball circuits apart thus loading opposite sides of the balls in each, and it’s possible to effectively create a double nut but tweaking the phase of one circuit within a multi-circuit nut. The brass nut in the video will wear before the balls, regardless of his assertions. It’s a shame that he didn’t try hard-milling those out of a better material – his machining was awesome. Kudos to anyone generating their own gcode from Excel!
I’m glad that he recognizes the limitations of his ball screw. As he said, a fun experiment, but even the cheapest ball screw is likely to be more precise and reliable.
I read all of that and I’m very familiar with the production methods involved.
You’re correct about whirling, and the best way to describe that type of cutter is an annular ring cutter that goes around the actual screw shaft, set off center of the actual screw stock bar. That ring rotates on a larger center offset from that bar and the cutters point from the inside of that ring toward the shaft. They have many cutting inserts on them annularly around the ring, and as the cutters go around the ring they are set into pockets slightly different distances from each other in a line, so the first cutter takes the least amount off as it rotates into the bar, and passes away at that apex point.
As the other cutters come around they progressively cut deeper in individual contacts as the major ring holding them rotates around, and effectively what it’s doing is threadmilling on a specialized form of lathe using an annular cutter with all the cutters pointing inward. There used to be manual machines that did this over a hundred years ago for the first precision lead screws in mass production, and the actual machine itself was called a Threadmill machine. It worked with an external cutter much like a rotary v-groove cutter, and travel down the bed of the lathe tilted at an angle the same way a modern thread whirling machine’s ring is tilted versus the axis of the screw to the pitch angle of the thread being milled as it travels down the rod on its tilted axis but aligned with the travel of the rod it’s milling.
You are correct about the nut wearing first here despite what I myself said above- only because in this case it’s higher pressures involved and they will swage the brass channels outward because he’s relieved the middle. But to a point, and as they are under pressure the balls will get worn from lapping action as well as micro grit gets in there. So there’s several things going on.
Thread whirling is actually how most bone screws are made too.
Thank you for taking the time to leave this comment, I learned a lot