It is surprising to find out that no one on the forum has ever mentioned a roller screw. :confused:

They seem to be a lot more robust than ball screws and look like something that could possibly be even made at home (a home equipped with a workshop and a lathe, that is :D )

Anyone ever pondered on the idea?

Posix,

That's different. First time I've heard of/seen one. Any more info?

Lance

It's pretty, though, ain't it? :D

Well, it works on the principle of planetary rollers - think along the lines of a rear diff on a car - the screw turns, little planetary rollers (smaller diameter screws with the same thread depth and pitch) roll in opposite direction (just like balls in a ball screw) around the circumference of the screw. The little planetary rollers also roll INSIDE a nut wich, also and predictably, has the same thread depth and pitch. So looking from the side at the whole concoction you in effect end up with a normal bearing with the shaft (screw) in the centre, balls or rollers (planetary roller screws) around the shaft and race (nut) on the outside.

I'll see if I can dig up a few drawings to make things easier to picture.

Ok, here's another couple of cut-away drawings. Also note that little (planetary) rollers have little pinions at both ends and the nut has a corresponding large toothed ring on each end. My guess is that's to discourage the rollers from "sliding" across the screw and nut surfaces but to turn properly. I've caught mention of gearing somewhere as well.

Could any of you solidworks gurus out there make up a solid model of this which could help us visualise the whole system. Also mechanical engineers among you could come up with a spreadsheed which could be used to calculate the exact screw, roller and nut diameters and thread depths and pitch of each component since this mechanism isn't using any standard threads. :eek:

I already had seen it in a (Hungarian?) Web.

http://www.hobbycnc.hu/CNC/Otletek/Otletek.htm

The Web is partially translated to English, but there is no explanation in English about the screw.

Somebody can translate it?

Hey, this is my first post!!!

I already had seen it in a (Hungarian?) Web.

http://www.hobbycnc.hu/CNC/Otletek/Otletek.htm

The Web is partially translated to English, but there is no explanation in English about the screw.

Somebody can translate it?

Hey, this is my first post!!!


nachini,

Can't help with the translating, but welcome aboard. Watch out-this site can be addictive! Enjoy...

Lance

http://www.exlar.com/products/linear/roller_screws/roller_screw_oview.html

Here is a web site with some good explanations...
http://www.exlar.com/products/linear/roller_screws/roller_screw_basics.html

It's definately neat. But it probably fits the typical model: if it looks complicated, it's expensive. If it looks expensive, it is expensive. If it looks expensive AND complicated...well like they say, if you have to ask the price you probably can't afford it. Bummer.

Lance

This is a good link http://www.exlar.com/products/linear/roller_screws/pdf/RolScrew12.pdf it's good some good pictures in it.

Looks like the standard nut has 0.01-0.03mm of backlash on it, but there is also a preloaded version. I am intrigued to know how it compares to something like this http://www.kerkmotion.com/ which claims zero maintenance with zero backlash, but there are no real values quoted, but does look a lot cheaper.

Stephen

An interesting point to note in the RolScrew12.pdf that Stephen gives the link for is that the machine tool applications that are mentioned are either high force or high speed but not high precision positioning. It could be that these screws can outperform ball screws in force and life but do not match them for accuracy and precision.

It would be interesting to try and build something of this nature using different sizes of threaded rod with the same pitch.

Well...just a quick glance and it appears to be less efficient then a ballscrew because of the increased number of parts that rub on the screw...okay, maybe they are suppose to rotate against the screw....it also has more moving parts that appear to require some precision in how they are mounted....those little planetary screws....they appear to float some....

I don't know but all material I've managed to come across lists roller screws as having superior accuracy and repeatability to ball screws and due to the large number of contact points also exhibits better stiffness and load resistance.

But all this is way beside the point. What I wanted this thread to concentrate on is making a DIY version of it as ball screws are nigh on impossible to make at home and these suckers look a lot less complicated.

And here are some examples of what people have allready done, albeit much simpler and cruder versions of what the real deal looks like:

http://www.hobbycnc.hu/CNC/Mech12/Zanya.jpg

http://www.hobbycnc.hu/CNC/Mech11/Csuszka1.jpg

http://www.hobbycnc.hu/CNC/Mech8/2.JPG

Well, hello everyone.

Today I had a productive day. For roughly $15 and in 15 minutes of cutting time I had my own roller screw. And I can tell you this thing beats any delrin nut any day. Ok it has some backlash, it does have some axial play as well but considering it's been made out of M10 "allread" and with only an idea in my head and no drawings of any kind, planning or anything else for that matter (even the cuts are as rough as they get - no polishing or finishing of any kind) this thing rolls really nice.

Now on thursday I'm going to put some plugs in my ears and go to the same guy with the drawings of a real roller screw and he told me that from what he made today and from what I explained to him about the real roller screws he could do something.

I'll try and upload a few pics of what I had made today. This is fantastic! Cheap homemade better-than-ballscrews roller screws are coming!!!! :wee:

I could envision putting a gear train on the rollers and driving that instead of the main screw as well, just look at that last picture and think gear driven.

Some images of what I've got:

P.S. sorry about the resolution; I have a feeling my digital ixus is going to pack up soon due to the lack of use... :frown:

All the home made designs seem to be missing the planetry gear that slaves all the screws together. I wonder what difference this makes?

Nice work BTW!

Stephen.

I haven't forgotten the planetary gear or the outer threaded shell but this was a test just to see how easy it turns and if the concept is ok.

The next phase is to take one of the cutaway drawings of a real roller screw together with all the elements and make the real mccoy, custom threaded rod, little threaded and toothed rollers, outer threaded shell, the works. Can't wait 'till thursday... :banana:

All the home made designs seem to be missing the planetry gear that slaves all the screws together. I wonder what difference this makes?

Nice work BTW!

Stephen.

If you look closely the gears DON'T tie the screws together-directly. They mesh with a ring gear. Here is my take. The little gears can't be of a larger diameter than the threaded piece they are attached to: the main screw would never fit though. But they could be the same or smaller diameter. If they were smaller, the ring gear would need to be resized to mesh properly. But changing the relative diameters affects he "gear ratio" between them. Thus, for a single turn of the body, you could get more or less rotation of the inner screws. If they spin faster, then the main lead screw will travel further for a given amount of nut rotation. So the ones without lose nothing-they are just operating at a "fixed" design point.

Posix, I posted earlier and said I thought this looked complicated/expensive. What you have done has changed my mind. I'm really impressed by the way you dove right in and tried it. Way to go!

Lance

I haven't forgotten the planetary gear or the outer threaded shell but this was a test just to see how easy it turns and if the concept is ok.

The next phase is to take one of the cutaway drawings of a real roller screw together with all the elements and make the real mccoy, custom threaded rod, little threaded and toothed rollers, outer threaded shell, the works. Can't wait 'till thursday... :banana:

Thursday, huh? You ambitious little devil, you. Question: how have you resolved the threads, pitches, etc? Inquiring minds want to know....

Lance

That's why I asked for the mechanical engineers among you to make some solidworks (or any other 3d) drawings and figure out all the dimensions. I have some simple and crude formulas for calculating diameters based on the drawings and photos I have so far.

The diameter of little rollers is ROUGHLY 34.09% of the screw diameter
the INSIDE diameter of the outer shell is roughly 163.18% of the screw diameter
thread depth on the screw is roughly 5% of its own the diameter
thread depth on one roller is roughly 14.66% of its own diameter
thread depth on (in?) the outer shell is roughly 3.06% of its own diameter.

So assuming a 16mm screw we have following:

screw ~16mm, thread 0.8mm, 5mm thread lead
roller ~5.45mm, thread 0.8mm, 5mm thread lead
shell ~26.11mm, thread 0.8mm, 5mm thread lead

Now, can someone with a bit of high-school trigonometry left in their memory confirm/deny this.

The DIY versions seems to run with the rollers only and the heavy duty versions have an outer "nut".
BUT, there also seems to be two different styles of production designs. One with the outer nut fixed as seen earlier in this thread, and one with the outer nut rotating, moving the axial load to the flange via roller bearings. Check the picture (or http://www.itwspiroid.com/rolrscrw.htm).

--Sven

....BUT, there also seems to be two different styles of production designs. One with the outer nut fixed as seen earlier in this thread, and one with the outer nut rotating, moving the axial load to the flange via roller bearings.... Sven

There is an additional difference:

The picture attached to post #4 has a caption that reads:

"Threaded rollers are the basis of SR/BR/TR/PR planetary toller screws."

But the link from post #22 has this description:

"What is a Spiracon® Roller Screw?
Our unique, patented Spiracon Roller Screw consists of a screw engaged by rollers which have annular grooves. Full line contact exists between the threads of the screw and the annular grooves of the roller."

For a DIY version it could be easier to do the annular roller version. It could remove the need to do precise measurements or calculations and also incorporate adjustment to remove backlash. A housing could be made to take six rollers with the bearings for the rollers axially adjustable. The adjustment is needed because annular rollers cannot all be in the same plane around the screw they have to follow the helix. Because the rollers do not have a thread but just grooves it would be fairly simple to machine samples to determine the particular diameter needed for a close fit. With six rollers all axially adjustable it would be possible to adjust three in one direction and three in the other to remove backlash.

In the same way that delrin anti-backlash nuts are designed, could the outer 'barrel' be made in the form of a delrin split tube which could be tightened down with a further sleeve/clamp.

Posix,

What are the conclusions so far?

Zoltan

SKF also makes roller screws. It looks like there are two designs, a Planetary setup (http://www.linearmotion.skf.com/SetLog/skfkey/ObjectKind/Solution/ObjectName/Planetary%20roller%20screws/ServiceName//url/_sl_Catalog_sl_ss.asp?FAM=SRrollerscrews_amp;Lang=EN_amp;ProfileName=US_amp;userId=0_amp;BackId1=121_amp;BackKind1=ProductRange) and a Recirculating design (http://www.linearmotion.skf.com/SetLog/skfkey/ObjectKind/Solution/ObjectName/Recirculating%20roller%20screws/ServiceName//url/_sl_Catalog_sl_ss.asp?FAM=SVrollerscrews_amp;Lang=EN_amp;ProfileName=US_amp;userId=0_amp;BackId1=121_amp;BackKind1=ProductRange).

Posix,

What are the conclusions so far?

Zoltan


It's very nice and as I've said earlier on - no delrin nut can come close to this. But bear in mind that I have used a normal cheap M10 threaded rod and the thread finish on that is, well, shall we say less than optimal :D

I will have a custom ground 16mm threaded shaft and little rollers and THEN we'll see what this beauty can REALLY do.

But for a first try this simple arrangement of 3 pieces of M10 "allthread" is REALLY a nice thing and I would wholeheartedly recommend to anyone starting out with cnc machines to go down this route if they cannot afford/find ball screws. It will only take half an hour for an experienced machinist to make you one and you could do far worse than substitute my steel retaining rings (the "washers" that hold the 3 pieces of M10 in place) with delrin equivalents for better rolling performance of the rollers. ;)

In the same way that delrin anti-backlash nuts are designed, could the outer 'barrel' be made in the form of a delrin split tube which could be tightened down with a further sleeve/clamp.

I think making anything apart from the positioning/retaining "washer" for the rollers out of delrin would actually be counter productive. You don't want anything sliding against anything else, you actually want things to engage other things and make them roll, just like in a ball bearing - nothing should ever slide inside a ball bearing but instead balls should engage inner and outer races and roll thus producing smooth motion. If sliding occurrs you end up damaging bits and getting less than optimal motion. So forget delrin, this should be all steel.

Point taken posix. Please forget I mentioned delrin. (chair)

What I was thinking was to make the outer sleeve adjustable in diameter in order to take up any backlash. Certainly I assumed that finely adjusted, this would allow the rollers to do their stuff properly, and not slide. Would this work with steel ?
I don't want to add any unnecessary bits, only remove problems. :D :D

It should all be steel and of course it would work perfectly. It's all in the maths required to calculate the diameters and dimensions of all parts. If you get that right everything works. If you are so worried about backlash then just put two of these babies back to back separated by a shim of some sort that would push them apart thus eliminating any backlash. But if everything is ground properly there should be no backlash to begin with. Even with my M10 "allthread" there is no PERCEIVABLE backlash - i.e. I cannot FEEL any on my fingertips but maybe a micrometer could measure some, I don't know.

Posix: Are you going to make a new one with a threaded nut, or are you going to refine the prototype you've made? I'm just curious. :)

I really like the first example you posted, no need for bearings and the backlash could easily be adjusted with the nut. I'm thinking that by having a "splitted nut", i.e. two nuts, the backlash could easily be removed by pressing them together by hand when fixating them to the axis. How about that?

--Sven

No refining, a proper one with threaded "nut" is next.
Yes, your picture shows the idea that I was thinking and that is your standard double preloaded nut. But you first have to build the one that I have to be able to appreciate just how little backlash this thing exhibits even though it has been made with most primitive of materials and techniques.

Can't wait to see your result! :)

Been googling a bit, really got this thing in my head now. :)

Found another nice design:
http://www.steinmeyer.com/english/technik/drs/drs_funk.htm (see the first image). No need for threading tools to make it.

http://www.about-ars.com/IHD/ARS_PS_RV1.htm and http://www.about-ars.com/IHD/ARS_PS_RVR1.htm describes the differences between satellite and recirculating rollers.
I know that SKF was mentioned earlier, but their PDF is very descriptive: http://skf.manager.nu/publication_files/pdf/7551246328.pdf (7,84 MB)
There's also a picture on the first side where the gear ring has the teeth outwards, would be easier to make at home. The recirculating version seems to be pretty straight forward, later on I'll try to make on of those or the "german design" I mentioned... :)

--S

You can google all you want but you won't find any numbers required to make them.

Well, here's one I made earlier and it really was an eye-opener and it also helped me fix some of my math.

Also I assumed only one start was required on the nut. Boy how wrong I was. You actually need 6!!! :D

You like? :rainfro:

P.S. all comments about thread collisions will be politely ignored untill someone tells me how to do patterns around a helix... :p

Good or bad news first?

Well, the bad news is that no one ever told me you can't just have any thread pitch you please. It has to be one of the preset pitches on the lathe. So my 4.8mm pitch for the screw was a no-go and a show stopper.

The good news is that after a brief discussion with the master I've come to the conclusion that it would be a simple procedure of re-drawing my whole assembly with a slightly shallower thread depth in order to get a standard pitch. So 0.5mm thread depth and 3mm pitch is on the drawing board.

For example I COULD go with a 5mm pitch on the shaft but that wouldn't give me a clean number for the thread depth (0.833333333333333333) and I would end up with a non-standard pitch on the rollers (1.666666667) which, again, is a no-go. You have to end up with the same thread depth and a standard pitch on all parts.

In a few days he'll give me a ring when he's cleared up his backlog so I can come over and we can tinker all night.

Is 0.5mm enough of a thread depth? Doesn't look that substantial in my drawings that's for sure.

Keep us posted, I carn't afford ye old ballscrews and I pondered on this one a while ago, I'm still in pondering mode LOL.

How about driving the nut instead of the shaft - long shafts may whip? if driven at the faster speeds needed? could you drive the nut in this design?.

Pat

Yes you can drive the nut and this design actually affords you higher speeds than a ballscrew.

I made some experimenting yesterday while watching TV. :)
The simplified "German version" should work with standard trapetz screws. Can't drive it with the nut and may not be as efficient as Posix's, but a lot better than a trapetz nut/screw. I'll try it out, just when I get some spare time... ;)

http://www.cnczone.com/forums/attachment.php?attachmentid=10707

Can someone buy me a lathe as an early christmass pressie? :cheers:

Nice thread! Looking at the drawings it seems you can take up any backlash by adjusting every other screw along the main screw axis. This might work better than trying to change the diameter of the outter ring or adding a second roller nut.

JR

Ok, today is the first day I was able to mount my roller screw assembly to my Y axis and compare it to the X axis.

Basically with the crummy "setup" I have at the moment I am able to squeze out additional 50% RPM more before the stepper stalls on the Y axis than on the non-roller screw X axis and they are identically built. AND the execution of my roller screw leaves a lot to be desired. Today I spoke to a different (and a lot older) metal master lathe person and he suggested some improvements on my original "design". And he'll do it even cheaper than the first bloke. Hopefully on saturday I'll have all 3 axes "roller screwed" and then we'll see.

I even spoke to him about making a real roller screw with all the little gears and things and he said it'll be difficult making the little gears with the machines he has (a big mill, a big thing that looks like a mill but has only Y and Z axes and a 1958 german-made lathe bearing a big "VDP" mark - anyone knows what it stands for?) but I told him I'm in no rush and that he can take as long as he wants with those. On saturday I'll take some printouts of what I've got to him for an oppinion. Fingers crossed my machine will fly on saturday! :bat:

Hi,

I have a question. I built one roller screw similar with yours, but despite the satellite screws are spinning freely when the main screw is not mounted, when I mount it and move it, the satellites are not spinning. Are the satellites of your construction spinning? Anyway, the screw is working more smooth and easy compared with the same screw with derlin nut, but I do not understand why the satellites are not spinning, despite not major friction is on their ends.

Thank you,

Zoltan

Hmm...if they are not spining in the opposite direction of the lead screw then it would be very difficult to turn because they would be sliding against the surface of the lead screw. In effect you would end up with just a plain nut and screw.

My rollers, or "sattelites" as you call them, are spinning in the opposite direction of the lead screw at all times, it doesn't matter if the lead screw assembly is mounted to an axis or not. If I hold the lead screw vertically and "push" the roller screw it will spin down for about 10cm before coming to a stop. It will spin up as well but not as long, only for about 3-4cm. I don't know what could be causing your rollers not to spin. Is your lead screw slightly bent perhaps? Maybe your rollers (sattelites) are not pushing against the lead screw tightly as they should? You have to calculate everything correctly. Otherwise it's :drowning:

For example I am using a normal M10 "allthread" rod. Looking from the front if the lead screw is in the middle the diameter around which lie the individual axis of rotation of the roller screws is cca. 18.5mm. How did I come at that number? Look-

M10 spec is:
outside diameter 10mm
diameter at bottom of tooth 8.376
diameter at the middle of tooth 9.02

9.02(roller on one side)+9.02(lead screw)+9.02(roller on the opposite side)=27.06 - this is diameter of the whole assembly
9.02/2=4.51 - this is the centre of a roller or its radius
27.06(whole assembly)-4.51(radius of one roller)-4.51(radius of second roller)=18.04

18 is too tight and you won't be able to assemble it. 18.5 is better - it is a bit loose and gives 0.23mm on each side but that number is good enough given that no M10 is exactly to spec since you always buy cheap iron bars with formed rather than ground threads with minor imperfections and 2/10ths of a mm is good enough.

Input your sizes into this calculation and see if your rollers are at a correct distance from the lead screw?

Hi,

Basically, I have the same figures, with one exception - 18 mm instead of 18.5, as I considered the medium diameter is 9 mm. How freely are your rollers on their bearings? Mines are rotating freely but not very easy.

Thank you,

Zoltan

It is surprising to find out that no one on the forum has ever mentioned a roller screw. :confused:

They seem to be a lot more robust than ball screws and look like something that could possibly be even made at home (a home equipped with a workshop and a lathe, that is :D )

Anyone ever pondered on the idea?

The idea isn't entirely new. It has popped up in technical magazines from the period 1950 - 1960 (can't say exactly when). Later it was fund having promlems with the contact surfaces due to very high pressure and a sliding movement of the contact surfaces. My opinion is that they are exellent in system that doesn't move very frequently, but of course I have been wrong before.

/jan

i find this interesting. how expensive are these things to buy anyway? Is this a case of spending all day to make a 5 cent screw? if it is inexpensive, it sounds fun to make ano matter what.

Hi,

Basically, I have the same figures, with one exception - 18 mm instead of 18.5, as I considered the medium diameter is 9 mm. How freely are your rollers on their bearings? Mines are rotating freely but not very easy.

Thank you,

Zoltan

That's because you went for 18 instead of 18.5. Mine are rotating completely freely, just like rollers in a bearing.

My opinion is that they are exellent in system that doesn't move very frequently, but of course I have been wrong before.

/jan

I don't think moving frequently is a problem, but sliding of the rollers could be. What if rollers gain a little more in friction in one direction and doesn't roll completely, instead making a tiny "sliding" movement? In this scenario the "gear ratio" of the assembly would change!

I'm thinking of making an assembly (when the time arrives, so to speak...) and make a test run, lets say moving the nut up and down on a screw a couple of hundred times and then see if it could go back to the starting position.

--Sven

I don't think moving frequently is a problem, but sliding of the rollers could be. What if rollers gain a little more in friction in one direction and doesn't roll completely, instead making a tiny "sliding" movement? In this scenario the "gear ratio" of the assembly would change!

I'm thinking of making an assembly (when the time arrives, so to speak...) and make a test run, lets say moving the nut up and down on a screw a couple of hundred times and then see if it could go back to the starting position.

--Sven

Sven in my limited grasp of this idea some of the designs - maybe the better ones for us - simpler - they use satellite rollers that don't actually have a thread form, but rather a series of grooves and ridges that match the thread form - kind of LOL. so a satellite roller slipping although quite undesirable in terms of wear will not alter any position.

now the difficult thing is that the inside of the 'nut' or outside case has internal grooves to mate with the rollers. those rollers or satellites must be of a size to pretty accurately prevent any 'slop'. that seems a little difficult to calculate/make?

am I grasping this design? or one of the types of this design :(

some designs will have simple gear teeth milled on one end of the rollers, those engage - I don't think it needs to be a particularly 'good' fit with internal teeth on the inside of the 'nut' hence making sure rollers cannot slip.

how about Delrin/acetal rollers - easy to broach teeth on?

wait :idea: internal thread on the nut - must use threaded rollers? otherwise it would jam. internal grooves must use grooved rollers. but surely the rollers cannot be allowed to slip and not roll... :drowning: heeelp

Pat

...Later it was fund having promlems with the contact surfaces due to very high pressure and a sliding movement of the contact surfaces...jan

The sliding of the satellite rollers is probably why they use the complicated sun and planet gear system with as many rollers as possible to distribute the high pressures over many contact points.

Pat; I think you have a good point "...- they use satellite rollers that don't actually have a thread form, but rather a series of grooves and ridges that match the thread form - kind of LOL. so a satellite roller slipping although quite undesirable in terms of wear will not alter any position."

I think for a low load application it should be possible to use this style and omit the outer nut. Mount the rollers in small thrust bearings to take the axial load. The bearings would have to be in adjustable mounts to space the rollers correctly around the helix.

I don't think moving frequently is a problem, but sliding of the rollers could be. What if rollers gain a little more in friction in one direction and doesn't roll completely, instead making a tiny "sliding" movement? In this scenario the "gear ratio" of the assembly would change!

I'm thinking of making an assembly (when the time arrives, so to speak...) and make a test run, lets say moving the nut up and down on a screw a couple of hundred times and then see if it could go back to the starting position.

--Sven

You got a point there... The "planets" can live e life of their own, separated from the rotation of the nut. I don't think I would use this design in a cnc design, at least not without a direct feedback of the position.
Still the problem with high contact pressure remains. Lubrication could be a problem as it is a sliding movement of the contact surfaces. In a ball-screw (or whatever the proper English word for it is...) the balls rolls into fresh lubricant all the time and also carry the load on a roling movement while the planetary nut pushes it away all the time as it slides.
/jan

:argue: (nuts) (flame2) (chair)
:tired:
:idea:

It seems that we have a bit of a language barrier here.

Nothing is supposed to slide on a roller screw assembly. That is the whole point of it. All surfaces roll against eachother in opposite directions. Lead screw rolls against rollers, rollers roll against the nut. No sliding.

Oh, and btw, there is no sliding in a roller screw. Nope, just rolling.

Did I mention that all elements in a roller screw roll against eachother rather than slide? Yep, rolling rolling rolling...

P.S. I almost forgot - lead screw and rollers roll against eachother, they don't slide. The rollers roll against the nut too. They never slide.
P.P.S. And one more thing - roller screws resemble a roller bearing in that the centre race - or lead screw - rolls and the little rollers roll around it. And the little rollers roll against the inside surfaces of the outside race - or nut.
P.P.P.S. Roller screws roll rather than slide. :cheers:

> P.P.P.S. Roller screws roll rather than slide.

I think no one had that thougt in any case, but how can you guarantee that they don't slide, even not a 0,01%?..
Of course they're supposed to roll, but I want make sure a DIY roller screw is useful.

--Sven

Visualising something without an actual model may prove to be very difficult so I urge you all to make one for yourselves. I have and my mind is made up. I wish you all made it at home and then we could all talk about ways of improving it from the same perspective rather than postulating and speculating on things that might not even pose a problem.

To directly answer your question about ensuring rollers roll rather than slide - make the "walls" of the roller screw assembly (the two pieces with little holes 120deg apart through which rollers go) out of delrin which will ensure smooth low-friction rolling of rollers around their axis and don't lubricate the lead screw as maximising friction between the lead screw and rollers is the goal to achieving rolling, as opposed to minimising it which would lead to sliding.

Gears help but can someone on here with home-shop tools make a 3.75mm diameter roller with 10 tooth on it and a 23mm ID nut with <insert the calculated number of teeth here> teeth to engage all rollers equally? Doubt it. I know I can't. So you make do with that you have. ;)

Hi, I have a quick question. I'm currently in the first semester of a tech program (manual & CNC) and my instructors think that I should go for an agressive project, like a manual bench lathe. I'm wanting to build one that I can implement at least a partial cnc system on (I already have a degree in electronics).

My question is: Would this roller-screw nut be appropriate for that application, or sheer overkill?

Thanks!

Re everyone making their own model etc... I'm going to try having the rollers turning on a single ball bearing at each end. Any use I make of the idea would only be for very very lightweight machining.
I hope to be able to drill a small dimple for the ball at each end of a short length of all-thread, as this would be easier for me to set up than turning down the ends.
Any comments ?

:argue: (nuts) (flame2) (chair)
:tired:
:idea:

It seems that we have a bit of a language barrier here.

Nothing is supposed to slide on a roller screw assembly. That is the whole point of it. All surfaces roll against eachother in opposite directions. Lead screw rolls against rollers, rollers roll against the nut. No sliding.

Oh, and btw, there is no sliding in a roller screw. Nope, just rolling.

Did I mention that all elements in a roller screw roll against eachother rather than slide? Yep, rolling rolling rolling...

P.S. I almost forgot - lead screw and rollers roll against eachother, they don't slide. The rollers roll against the nut too. They never slide.
P.P.S. And one more thing - roller screws resemble a roller bearing in that the centre race - or lead screw - rolls and the little rollers roll around it. And the little rollers roll against the inside surfaces of the outside race - or nut.
P.P.P.S. Roller screws roll rather than slide. :cheers:

Ok, ok, I wasn't thinking of sliding like in a solid brass nut, I was merely thinking of two convex surfaces in contact under high pressure. The screws with balls are a little better of as a convex surface meets a concave. Of course I wasn't thinking of sliding like a screw in a solid brass nut, but let's leave it. The contact pressure is still high as the surfaces in contact vitually have a 0 area, if not, they must slide...
Never mind, If anyone can get this device to last long it's ok with me. I see it merely as a component most useful in devices with a limited expected lifetime like in a jack that only would be used a few times but where high forces are expected at minimum weight, minimum volume and low cost. On the other hand, as I stated somewhere else at cnczone, I might be wrong. It wouldn't be the first time and probably not the last.
And - yes, I have a problem with the English language but I'm working on it. I'm how ever jolly good at Swedish :-)
By the way, I'm a test engineer by profession and looking for possible problems have became a bad habit. Sometimes i might find a problem that not really is a problem.
:cheers:
/jan

The problem with using threaded planets is one of repeatability. It's like you're driving your stage with a friction wheel, yes it rolls, but there's going to be some positional losses. Just find two same size screws and play with it for half an hour, that's what I had to do to visualize this. Now if instead there were just stacked consecutive rings of teeth with the same pitch as the drive screw, there wouldn't be a positional loss problem, but it would be highly efficient since the planets are making rolling contact like a ballscrew. If you made a similarly designed ring roller, there would be no need for a complex planet retainer with thrust bearings and such.

Unfortunately making the planets and ring roller would require a lathe and some skill. There wouldn't be the benefit of creating gear ratios either. Despite these shortcomings this could still be an inexpensive alternative to ballscrews even if someone were to have a machinist make these components.

It's perhaps worth looking at a rather large pdf file from SKF on roller screws, I carn't seem to get to 'official' site link to produce the file, however I found the link from the Hungarian web site someone posted early in this thread.

http://www.hobbycnc.hu/CNC/Otletek/Gorgos_anya.pdf

it's a 5.4Mb file mind

<snip>
Gears help but can someone on here with home-shop tools make a 3.75mm diameter roller with 10 tooth on it and a 23mm ID nut with <insert the calculated number of teeth here> teeth to engage all rollers equally? Doubt it. I know I can't. So you make do with that you have. ;)

Posix, I don't know if the rollers have to be so small, but then again if you want six or so... maybe we are back to looking at less? keeping them reasonably long still gives a lot of contact and hence load carrying capability?

ok if we make them big enough it shouldn't be too difficult IMHO (ok I have some access to a small worksop at work) to perhaps broach those teeth on?

I'm reckoning that no great precision is needed just the requirement that the rollers must not be allowed to slip very far - maybe I'm kinda guessing wrong there?

well particularly if you can use Delrin it broaches real sweetly !! the internal gear teeth - I don't think that the number of teeth would be important? on some of the piccys that internally toothed ring looks like a seperate part - might be best me thinks!

Now I was dreaming the other day of how to use a single point broach to broach multiple teeth around a diameter without using a rotary table. on a 3 axis CNC mill. I must have a go sometime... take too much to try and describe will maybe try to explain on another thread.. if time permits

I have my set of 3 roller screws. But I think that the decision to go for a cheap off-the-shelf M10 "allthread" was a wrong one. I should've asked for a steel M10 allthread to be made on a lathe for me. As it is the roller screw is a bit dissapointing as the cheap allthread isn't exactly to tolerances so it wobbles a bit as the space between rollers and lead screw isn't the same throughout its travel but fluctuates somewhat. Serves me right for not having take this into consideration before I had these made. But it would now be a simple matter of having a precision M10 rod made and then use that for the rollers. The rest can be reused. I think.

Hmm... I have a set of three ballscrews without nuts, maybe should convert 'em to rollers?..

You can't as the ball screw shaft is a "female" pattern -v-v-v. You would have to make a "male" roller -^-^-^ and that would be too much effort to be honest. You need something "unisex" - i.e. it has to be the same on both sides. And the only thing that fits the bill is a seesaw thread like so /\/\/\/\

Much easier to just find a piece of hard rod and start from that.

I have my set of 3 roller screws. But I think that the decision to go for a cheap off-the-shelf M10 "allthread" was a wrong one. I should've asked for a steel M10 allthread to be made on a lathe for me. As it is the roller screw is a bit dissapointing as the cheap allthread isn't exactly to tolerances so it wobbles a bit as the space between rollers and lead screw isn't the same throughout its travel but fluctuates somewhat. Serves me right for not having take this into consideration before I had these made. But it would now be a simple matter of having a precision M10 rod made and then use that for the rollers. The rest can be reused. I think.

Thats a bit sad... the threaded shaft must be consistant throughout it's length.. you've got me wondering how they make that 'studding' I believe though I'm not sure... anyone?... that it would usually be rolled and I'm thinking does that produce a thred that is consistant throughout it's length? if you only want a shortish length then matbe you can get one turned but be careful if it's screwcut without any support on the sharft then that would cause it's own problems.

I was wondering about using Stainless steel - do they still roll that stuff?

Well now you mention the supports when I spoke to this chap about making me a threaded rod with my spec thread depth, angle etc he immediately asked how long and that it should be done with supports - I don't know their proper name but lathe people will know what I'm talking about.

And you know what the saddest part is? All that time I had a 1m piece of 20mm hardened steel rod in my car! :(

Svenakela, I have the same problem. I just bought three lengths of 1 inch precision ground ballscrew, but the nuts are all missing! I did figure out a way this could be done with them, though. If you have plenty of screw (as I do), you could cut three small lengths of the parent screw to make each nut. You then have only to fill the tracks with something round (surface ground mild steel or brass) in the proper cross section on the satellite rollers, and you would have an efficient set of mating rollers. Due to the large surface area, I would probably be able to get away with using 1.5 inch satellites on my 1 inch diameter screw.

Made a model of a recirculating nut (i think this is the one most suited for diy)! Seem very doable! Have to try it :) The problem with this design is that you can't use it as a rotating nut. But i guess you just have to make the spacers of metal and connect them to the housing in some way.

I belive you could get very good results if you use a threadmill and a rotary table on a mill! Easier than doing it on a lathe, and you could make smaller rollers and ID of the housing!

/Jay

Corvus
Any chance of finding springs of the right cross section that would grip your "rollers" as thread fillers ?

Corvus: Isn't irritating, the lack of nuts? ;)
I was thinking of making a couple of rollers that are not threaded, just with grooves, to run with the ballscrews (nope, not hard to do if someone you know have a CNC-lathe). Something like the german design I've mentioned earlier in this thread. My main question still needs an answer: What happens if the rollers stops? Even though the rollers are not threaded the gearing will change and I don't want a "dynamic gear ratio"... ;)
I was talking to a workshop that makes threaded rods, and HOLY CRAP there was a big difference in price when I questioned for better precision. That's why Im going to test my ballscrews, I already have them and they are useless...

JBV: Very nice assembly! Sure you can use it as a rotating nut, why shouldn't it work? As Posix has mentioned, the rollers are supposed to roll, and then it doesn't matter if the nut is rotating or the screw. The relative movement between the parts is the thing that counts.

Cheers,
S

Yeah but in my assembly the spacers are made of nylon or some other plastic and not connected to the housing, and if you turn the housing they almost definitively would slide :) You have to transfer the rotation of the nut to the planetary rotation of the rollers without "losses" if it's going to work predictable and with any degree of accuracy.

Should be easy to integrate the spacer in the housing (and endcap). I'll try to make a more advanced model! Think it would be a good idea to have wipers on the nut?

/Jay

What about using the same kind of leadscrew as rollers and just turn them down and the ends with a groove?

The housing is a bit harder to make but i belive you could make it in separate pieces and bolt together. Perhaps with some kind of preloading of the ridges the rollers rides on :)

Suggestions for the spacer? I have some ideas but they are hard to assemble!

/Jay

Possibly, greybeard- I still have'nt recieved the screws yet, but when I do , I intend to measure them to find out what size they really are- I'm pretty sure they are 25mm X 5mm lead, not 1 inch by 5 thread as the seller thought. That would be a very unusual size- I should know, my Index mill had 1 inch by 5 Acme screws in it, and when I had to rebuild it, I had to have them made, at about six times the price. :( I will probably just buy some nuts made for rolled screw of the same size, and hope for the best. As it sits, it looks as if these screws are going to spend some time on the shelf, next to all the other worthless parts I've accumulated in this project. Having to buy surplus kind of sucks. Its an EXPENSIVE education, but no different than any of the other five hundred other subjects I've become expert in- by doing everything WRONG the first time.

If the rollers and nut are grooved instead of threaded, it won't matter if they slip or not, your screw will advance the same amount, just not as efficiently as if they were rolling. Like JBV's first model, the second would have problems with variable gearing.

http://www.linearmotion.skf.com/en/0/products.htm
----------------------------------------------------

Here's a really detailed document on "Roller Screws" by SKF

http://skf.manager.nu/publication_files/pdf/7551246328.pdf

I'm not to sure I understand the difference between Recirculating and Planitary type roller screws? Both are high dollar!

Eric

http://www.linearmotion.skf.com/en/0/products.htm
----------------------------------------------------

Here's a really detailed document on "Roller Screws" by SKF

http://skf.manager.nu/publication_files/pdf/7551246328.pdf

I'm not to sure I understand the difference between Recirculating and Planitary type roller screws? Both are high dollar!

Eric



The planetary roller screw has threaded rollers that not only rolls by themseleves, they also roll around the screw (like planets) and are "geared" in the end to have a fixed rotational speed.
The Recirculating screw roller has grooved rollers and the roll freely and are only forced to have their fixed position by the spacer that keeps them from each other. This version also has smaller resolutions because of the "lack" of threads.

--S

Corvus - "Its an EXPENSIVE education" - I attend the same college !
Widgetmaster - that is some link to skn roller bearings. It's going to take a week to read, but I guess it's all there.

hi guys! have any more simple roller screw to build?

from http://www.nookindustries.com/RollerScrew/RollerScrewGlossary.cfm:

"The rollers have a gear and a cylindrical journal at each end. The gear at the roller ends meshes with the ring gear (4) at each end of the nut. This mesh prevents unwanted roller skewing by maintaining parallel axes of the planetary roller and the screw shaft."

Is the only function of the gears to prevent the skewing? Couldn't you solve that by putting the planetary rollers in a frame instead? They can't skew at all then and no need for timing. (you would have to put in the frame into the nut first and then one roller at a time and press the roller pin trough the small holes)

That's exactly why there is a planetary rolling version and a recirculating version. You just described the recirculating one. :)

--S

But the biggest difference is in the threading versus grooving!? Planetary have threaded rollers and nut, and recirculating have groved. I just think those gears seem unnecessary complex if its JUST to keep the rollers paralell to the leadscrew :D

What i try to say is, if you use threaded rollers you must have multistart thread (as many as you have rollers) in the assembly, the nut lines up all rollers and grooves perfecly, and would not mesh with a unmatched leadscrew. That rules out any simple diy versions.

The planetary version could use ordinary threaded bar for both rollers and leadscrew but the gears in is to hard to make, so i want to get around them someway! Is that possible?

I have an ideay of making grooved rollers and a very simple nut (even more simple the German one) where the screw can be single threaded. The bad thing is that the work load on the rollers will be higher and the nut will only take the load at the ends.

Lets say I'm using a thread with 5 mm climb per rev, 5 grooved rollers needs to be moved 1 mm axially to each other. Simply it can be made with 1 mm bushings. If it's for a machine with moderate loads, it could work. Se my very simple picture.

It would also be possible to make a German version-assembly, but then the grooves on the rollers have to be placed differently. On the other hand, if the rollers are going to be made in a workshop lathe it doesn't really matter, they are all made separately.
So my conclusion is: It should be possible to make a roller screw with single threads.

--Sven

I have an ideay of making grooved rollers and a very simple nut (even more simple the German one) where the screw can be single threaded. The bad thing is that the work load on the rollers will be higher and the nut will only take the load at the ends.

Lets say I'm using a thread with 5 mm climb per rev, 5 grooved rollers needs to be moved 1 mm axially to each other. Simply it can be made with 1 mm bushings. If it's for a machine with moderate loads, it could work. Se my very simple picture.

It would also be possible to make a German version-assembly, but then the grooves on the rollers have to be placed differently. On the other hand, if the rollers are going to be made in a workshop lathe it doesn't really matter, they are all made separately.
So my conclusion is: It should be possible to make a roller screw with single threads.

--Sven

At the bottom of post#23 this is what I was suggesting without a picture. I suppose I should learn how to draw so I can include pictures. I also added that it might be possible to include backlash adjustment.

Check this out regarding the "german variant": http://www.steinmeyer.com/english/kataloge/pdf/drs.pdf

On page 6: "The rollers work under "rolling friction" to the screw and to the nut. This fact causes a certain lost motion. Therefore there is no repeatability of
the feed. In case of repeatability is required, we are in need of an additional
measuring system! On most applications, two end switches will
solve that problem."

Seem you would need an absolute linear encoder to use these! :P

In your variant Sven, do you think small missalignments would prehaps remove backlash as a bonus? :)

Ballscrews actually have the same principal, they roll with friction. I don't see that the rolling through friction idea forces me to use a linear encoder. As long as the friction is correct adjusted it should work fine.

I believe backlash is gone! :)

--S

Hello...
Check...
http://www.hobbycnc.hu/CNC/Mech8/Mechanika8.htm

From picture the rollers seems to have groves only (not thread) and does not rol around main shaft. Just rolling groves. That seems to be just like better nut. If sometime slips there is NO gearing efect just a litle more friction I think.

Hey, that looks nice, and if those rollers are grooved instead of threaded, there wouldn't be any positional losses like you said! Too bad a lathe is required to make those rollers, but it's definitely alot simpler of a design than what others have been posting, plus it's fully adjustable so you don't have to have perfect tolerances. I like it.

Hello there peoples... Long time no post :rainfro:

Anyways, I talked to some chaps at a tooling dept of a local (to me) metal working company.

They tried to make something for me and it turned out that 4732 wasn't stiff enough for what I had designed. Also it's going to be very difficult persuading someone to make you a non-standard thread shape (and I believe 90deg or even shallower is what they use on roller screws).

I've tried an M14 lead screw with 8mm "sattelites" with the same thread profile as M14. Basically an 8mm sattelite isn't tough enough and it bends on the lathe just as you try to do the final pass of the thread. So I'll either have to find a MUCH stronger material (even though 4732 should've been strong enough) or re-design with something thicker, like 10mm or so. But then the whole lot becomes very bulky indeed.

Anyone in the metals industry fancy taking a stab at guessing what materials SKF or whoever might use for their roller screws? All I know is it looks shiny but nothing other than that.

Did you use M14 fine or coarse? The look of the SKF stuff seem to be more like fine :) Parts is also is precision ground (of pretty hard steel i would guess :))

/Jay

"pretty hard steel" won't cut it this time, though :o I need names and numbers.

And skf stuff is coarse thread rather than fine and even though it may look like it it's 3 (or even 5) coarse ones packed together onto a single lead screw. And they don't seem to be standard 60 deg threads either but more like shallow 90 deg jobbies.

I have been thinking of doing some tests with the simpler variants like in the Hungarian? page... M20/M16 Fine as leadscrew and M10 Coarse as rollers! They have 1,5 mm pitch so they should mesh just fine! There is room for up to 8 rollers around the (M20) leadscrew...

Do you think it would remove backlash if i used long rollers? Like 10cm or so :)

/Jay

It's the way they mesh that is vital to removing backlash rather than the length of rollers so it's all in your thread profile really.

M16 fine and M10 coarse is an idea but you should really think about sizes of the finished thing as well. But it's worth a try anway :idea:

Posix - sorry to be a pain, but could you point me to a drawing or description of the arrangement of the rollers/screw that your going for ? I've gone back through the thred but I'm not sure of your current thinking.
Jay - are you following the same path, or a variation ? I can't spot the "Hungarian" reference.

John

Have a look at those photos at the top of 2nd page in thread. That's my current "prototype" setup. Basically an arrangement of 3 M10 "allthread" rollers at 120deg apart around a central M10 lead screw. I'm working on something more elaborate (steel lead screws and rollers being made as we speak) with lots of drawings and things.

I was thinking that if you have longer rollers, the pitch errors in the rollers makes the meshing tighter than normal in some places (and ofcourse looser in some)...

This is the page i was refering to: http://www.hobbycnc.hu/CNC/Otletek/Otletek.htm

Thanks guys, I'm up-to-date now.

re backlash in general. If the three rollers have a small degree of elasticity, ie are made from something like delrin/hdpe, and are 'preloaded' by adjusting their position end to end with reference to each other, would this be a suitable way of reaching a compromise between backlash and friction ?

Imho adding to the length will add highs and lows in proportion to length. Reducing the length to a minimum might be argued to be a better approach.

I've been struggling :rolleyes: with how to get three single balls to replace the three rollers, having aquired some 2mm ball bearings. Initially I was thinking of building a ball nut of my own design to roll round M16 studding, but then you started this thread !

I've just found those ball bearings, and measured them - they are only 1.5mm.
The little grey cells clicked, and I'll try and draw this up tonight, but for now -

M16 threaded leadscrew
thin annular ring of phosphor bronze retaining three 1.5mm diam balls, each at 120 degrees and offset to sit in the thread
outer sleeve with three grooves inside, each ball running in its own groove.

Possibly duplicate the ball/groove set up with arrangement for 'preloading'.

I've no idea how much torque this could transmit, but I have the balls/M16 studding, but so far no idea how to make the outer sleeve without a lathe set up.

That's a ball screw. Get out of this thread! :)

No,no......(cries of innocence against hard-hearted posting)... I've just shrunk the rollers dear.

But anyway there would be so much friction between retaining rings and balls (balls would slide against walls of holes on retaining rings) that it wouldn't really be usefull as a proper ball or roller screw and you would esentially end up with just a screw and nut combo. Think no friction. Think rolling. Think no sliding. :wave:

OK Boss :idea:
(exits stage left with clean drawing board, a single ice skate, and a mobius strip made of teflon)

Did you get far along that strip yet?

Started to worry when I met myself coming the other way............

Me too................... worried that is lol, Have been watching this thread since joining this community, can whole heartedly say, very interestiong ideas here :) I need these roller screws (homemade ones) ................

Dunno how I going to make em thou lol, not lathes no machining facilities etc, all by hand arggggggggggggggggggggg.

Ideas fronm you folks thou makes me think much, not sure thats such a good idea thou lolol.

Mike

I've always passed this post by, but today I decided to take a look. I didn't follow all the links and read all the PDFs but it seems to me that you could combine the original prototypes/trials by POSIX (pictured in posts #13 & 16) with JBV's design as shown in post #79.

:confused: Why can't you have two rings of 3 rollers that alternate with each other? The rollers on the inner ring would be threaded and "ROLL" to match the All Thread/ACME Threaded lead screw (POSIX) while the outer ring would be comprised of solid rollers that would "ROLL" in a groove(s) (JBV) in the outer housing. The diameter of the outer ring would only have to be slightly larger than the inner ring. Just enough to keep the threaded rod away from the outer housing and the solid rollers away from the lead screw in the center.

Does this sound plausible? Do-able? Am I missing some key concept?

Can someone with better (okay, faster) design skills mock up a drawing? This seems like it would be easier to implement for the DIYer as it doesn't involve bizarre thread sizes with multiple starts, etc.

Just a thought,

HayTay

Yo yo yo people!!!

I got some updates coming this way soon!!! AND GREAT UPDATES!

I have been at the machine shop (umpteenth by now) and have seen the first of three roller screws that they made for me.

All I can say is *THUD*


Erm...someone dropped an anvil on my head - I guess I'm supposed to keep quiet about it untill it's all done and dusted...

Can you wait till thursday

I promise this will be a cracker!

Thursday's come and gone, we're waiting with bated breath!! lol

Posix

We are looking forward to seeing what you had made. We will be here when you're ready.

I was trying to avoid this thread due to possible vigelante attacks and excessive amounts of violence that loomed upon me...

Oh well, it looked promising. Then I asked to do one more. Then everything went downhill from there....

I'm going back to the shop tomorrow with a bigger rod.

Basically, their sliding thingy couldn't cope with a 12mm rod that I wanted to make a screw from. And the lead screw came out bent, threads were misaligned (we are talking 10s or several 100ths of a mm), rollers were out of position (a 10th here, 10th there) etc etc. So I just broke down and couldn't take it anymore.

The chaps at the shop asked me if we could give it another go this time with a bigger radius rod so that's what we'll do. 20mm instead of 12 this time. And 9mm pitch. This should be fun. I will get back with pics as soon as I've got something working.

It sort of showed promise but then dissapointed big time. And it was all because of stupid fractions of a fraction of a millimetre! You know the feeling - when you're on this high, when you have a hunch that everything will work out and then everything falls apart because of a stupid 100th! GRRR!!!!

They made me a teflon nut for one of the lead screws they made for me to take home and play with till tomorrow (4.5mm pitch, 3 starts) and, albeit with some binding (remember, threads were off ever so slightly?), this thing really flies with my puny steppers! Imagine what a 9mm pitch, 5 starts, roller screw should be able to do? Wheeewwww.....

Oh and BTW - Delrin DEFINITELY CANNOT take any loads of any kind! Full stop!

I would also ask any metal guys out there to recommend an alloy for the roller screw assembly. We tried stainless steel (could it be 430?) and even though it looks pretty I have a feeling it's on the soft side. One of the numerous tool steels perhaps? 1530? 4532?

Why do I have an eerie feeling that everyone hates me right now?

Although I have nothing to contribute to your questions, I by no means hate you, and respect all the time and effort you've put into this device! :) While time is not completely on my side, I do have the patience and confidence to wait for what you come up with.

I, for one, will patiently wait until you have something to show. Thank you for all the time you have put into this project.

we are all waiting like vultures over a dehydrating and delerious posix...... will he make it to the watering hole or will it be ...... just kidding and waiting to celebrate with you on a major accomplishment. no matter how much it took to get there once you have arrived, your there!

Amen from me too. Lots of luck.

In the interim of posix getting his thing done, I present unto you a roller screw design I've been working on, and plan on prototyping if I ever come up with a CNC frame I like (don't hold your breath). While it looks complex at first, I tried to design it from the aspect of as little complex machining as possible. The three "planets" around the central 1/2-13 UNC rod use a straight grooving instead of a thread/helix, so that there aren't positional errors due to minute slipping between the planets and central screw. Of course, you don't get the speed multiplication that threaded planets provide. The planets have inlaid bearings (cheap from VXB) which ride on 1/4" rod. I couldn't find inexpensive thrust bearings, so I figure take a pair of nice metric washers (the bore on an M6 washer is a perfect fit for 1/4 rod) and sandwich some small BB's, while fabricating a retaining ring out of thin sheet metal, ideally brass. Finally sandwich it all between some angle and it's good to go. If the holes in the angle supports are oversized, it should be possible to bring the planets into contact with the screw with an additional light preload.

Hmm...zach_g you have no idea how much grief you've caused me with the statement "straight grooving instead of a thread/helix, so that there aren't positional errors due to minute slipping between the planets and central screw. Of course, you don't get the speed multiplication that threaded planets provide."

I have completely forgotten to think this through. I mean, the rollers in my mechanism roll even without pinions etc but even the slightest possibility of them slipping induses a positioning error just because the roller is threaded and no longer runs at a predefined ratio to the leadscrew but instead is standing still and slipping which makes the whole mechanism run slower (for the pitch of the roller per revolution). And that's a big issue!

Only now am I beginning to understand why is there a planetary grooved version as well - it IS slower but slippage doesn't matter.

Ugh!

Sorry to hear that. Not to come off as rude, but the problem of positional error due to slip in threaded planets has been brought up alot previously in this thread.

Maybe you could adapt your design to use straight grooves? If you had a 3 start screw you could arrange 3 planets inline with each other and use a straight grooved housing. 4 start could do 4 planets etc. Or maybe have a single start thread, and groove your housing with a pitch 1/3 that of the rollers or space out the grooves on the planets and housing and have each planet offset from the others.

While my posted design might be easier to make, it has a high part count. Yours is definitely simpler in that respect, and I hope it works out! You could always use a linear encoder too.

hey zack_g
howdy!
how many rollers could you fit on a 1" shaft? with whatever thread, course or fine would give you near 20 deg. of pitch ? and while you are at it lets make the assembly fit into a hollow center pulley so i can drive it 10' up and down my fantasy plasma table(sheet steel is 4'x10' most of the time on the turd coast). man it would be sweet to hold the shaft stationary and drive the gantry from a motor and pulley arrangement! oh and how about make half of the rollers shafts 1/8" longer with a washer and split nut to take out any extra clearences that might arise from wear. can you visualize the things i mentioned?

You could probably fit 6 rollers around a 1" shaft, but 3 is all you need really. There are plenty of points of contact for load handling. Using a wave washer would be a good idea to compensate for wear, although you could never apply a load larger than the spring force of the wave washer, which would limit the acceleration on a plasma table. Mount the whole thing between 2 big angular contact bearings and you should be good to go. I'm sure it will be a lot of fun to align everything concentric, of course if you already had a machine you could cnc some accurate plates to hold it together.

As far as what screw would give 20 deg pitch, isn't pitch usually measured in length between adjacent thread teeth?

im not knowing of the pitch question, could the angle iron be converted to a round shape(or pipe) that would fit between a pulley and the 1"shaft? would side loading from the motor torque and belt tension create drag and further slow the velocity? would reversing the direction compress the wave washer? etc?etc? ad infinum.. too much conjecture occupies the mind. remember to swallow!

I have been thinking of try to build something like this for my mill... There is very little height space for the X axis nut (23 mm) but my design is only 22 mm for a 16 mm leadscrew :)

1: Spacers/washers of different thickness to line up the rollers with the leadscrew.
2: Lock nuts to thigten the rollers so they can roll easily but not move axially.
3: Brass bushings.
4: Washer.

Think it would be possible to turn the rollers (10 mm diameter) on the mill using a endmill holder and a lathe tool held in a vise?

/Jay

Whenever I think about the "sliding and gearing problem" I forget to take one thing into account:

Does it actually matter if one roller slides in a 3-roller arrangement when you've got 2 other rollers that are still rolling? I know rolling resistance will grow. But will the gearing be affected?

Can someone elaborate on this a bit further?

The problem with the threaded/helical planet roller screw is that it doesn't have quantized positional steps as compared to a grooved planet. The threaded planet rollerscrew relies on rolling and the interlocking of the teeth on planets and screw to advance on the screw, while the grooved planet only relies on the interlocking between the planets and screw. The rolling of the grooved planets simply reduces friction, and doesn't advance the rollerscrew. The analogy would be using a friction wheel or a rack and pinion to drive a table. While both would position the table well, only the rack and pinion would have repeatability in position over a long period of time.

Well I would beg to differ.

Let's have an example here:

20mm threaded lead screw.
9mm lead, 6 starts, 1.5mm thread profile

8mm threaded roller
1.5mm lead, single start

Now, for every single revolution of the lead screw the roller has to make 6 revolutions in opposite direction to keep up "on thread"

Can a physicist or a mech engineer among you come up with some formulas to calculate the advance the whole assembly would be making for a single revolution of the lead screw. I know that if all rollers were sticking it would be 9mm. But what with rolling rollers?

I know that in some of my simulations on a 3-start 12mm 4.5mm pitch with 1.5mm thread profile leadscrew the actual distance travelled was 10mm for every revolution. I don't know how I came by that number but it seemed to be the only number where everything would interlock properly. All others were off somehow. I don't even see what relationship that number has to anything apart from being:

lead screw pitch+(roller pitch*number of revolutions of roller for every leadscrew revolution)

But that's not the correct answer I'm sure. Can't be because the roller rolls in the opposite direction so that would come out at -1 rather than 10.

Well I would beg to differ.

Let's have an example here:

20mm threaded lead screw.
9mm lead, 6 starts, 1.5mm thread profile

8mm threaded roller
1.5mm lead, single start

Now, for every single revolution of the lead screw the roller has to make 6 revolutions in opposite direction to keep up "on thread" ....

I think your last sentence has an error; The lead screw is 20 mm, the roller is 8mm so if slipping does not occur the roller turns 20/8 = 2.5 rev for every rev of the lead screw. Actually it will be a little different because the screw and roller make contact at some sort of pitch diameter which is a bit less than the O.D.

So for every lead screw revolution forward you move 9mm but the rollers have turned 2.5 rev back which is 3.75 mm so your net movement is 5.25 mm.

With the threaded rollers it is important that all turn at the same rate hence the fancy planetary gear train.

With parallel grooved rollers they do not have to turn at the same rate because the "lead" is zero. Actually if you locked all the rollers it would be a regular nut. So with these your motion is whatever the lead on the lead screw is.

Are you 100% on this: "the roller turns 20/8 = 2.5 rev for every rev of the lead screw". I can understand this for a single start thread on both pieces but this time we have a multiple-start threads on the lead screw and a single start on the roller.

The lead screw has a 9mm thread pitch and the rollers are 1.5mm so can you imagine the way their respective threads interlock and the resulting rotations? I can but only with the help of some motion animation.

Ok I've tried to simulate your scenario and OF COURSE it doesn't work [:D]

But what DOES work is that you actually add up the values 9 and 2.5, then it works.

As I said (on numerous occasions before) it is very difficult to visualise the way this thing works without either looking at it in flesh or having it on screen turning.

Here is a quick animation for you (divx) :cheers:

If no slippage occurs the two parts rolling against each other have to turn in the inverse ratio of their diameters. I don't see how it can be any other way; they have to maintain the same peripheral speed at the point of contact.

I think the multistart thread on the lead screw is just needed to keep the helix angle the same at the pitch diameter.

But I did have a mistake; the rollers rotate in the opposite direction to the lead screw but the two leads do not subtract they add so the net relative motion between the lead screw and nut assembly is 9 + 2.5mm = 11.5mm

Ok, fine. But now what happens if one is slipping but the remaining 2 are still rolling? Or if 2 are slipping but the 3rd is still rolling? What then?

I guess what I'm after here is for someone to tell me that there won't be any positional errors if slippage occurs and as long as at least one roller is rolling.

Or put another way - this might still work without the pinion arrangement.

There will be positional errors if slippage occurs. Think about a single roller situation: when both screw and roller are rolling the net relative motion is the sum of their individual leads; if the roller stops the net relative motion is the lead of the screw. Now move to multiple rollers; when one slips a little it starts to fall behind the others but they are in a fixed assembly and all have to move axially at the same rate therefore binding occurs.

As I mentioned earlier with parallel groove rollers the net relative motion is the same whether the roller is rolling or fixed so with these some slippage does not escalate into a binding situation.

The slippage being referred to isn't where one roller might lock up, it's the minute shifting at the point of contact that all the rollers exhibit while rolling. It's inevitable that there will be positional errors. If you don't believe me, believe Steinmeyer in their pdf on roller screws http://www.steinmeyer.com/english/kataloge/pdf/drs.pdf

"The rollers work under "rolling friction" to the screw and to the nut. This fact causes a certain lost motion. Therefore there is no repeatability of the feed. In case of repeatability is required, we are in need of an additional measuring system! On most applications, two end switches will solve that problem."

Sorry, I know it sucks to have your design that you've put alot of effort into not do what you thought it would do. All is not lost though, you can still modify your design to work, I posted a few ways earlier. The key is to use grooved planets instead of threaded, that way the lead won't depend on rolling friction, but you still get the efficiency of rolling friction like in a ballscrew. I think the binding issue Geof presented is also a problem, unless there was a retaining ring that kept all the planets equally spaced like you see in alot of commercial varieties.

I think your last sentence has an error; The lead screw is 20 mm, the roller is 8mm so if slipping does not occur the roller turns 20/8 = 2.5 rev for every rev of the lead screw. Actually it will be a little different because the screw and roller make contact at some sort of pitch diameter which is a bit less than the O.D.

So for every lead screw revolution forward you move 9mm but the rollers have turned 2.5 rev back which is 3.75 mm so your net movement is 5.25 mm.


Wow, someone agrees with what I pointed out. Gear reduction occurs when rollers are allowed to roll.

Some thoughts. 1/100 of a millimeter is getting pretty small for deviations (1/10 is getting p there, but still only .004"). Lapping compound with a run in period and then a final cleaning and lubing will normalize all/most inconsistancies. Posix you had one of your attemps traveling less than you thought it should, but with the gear reduction, it might have been as it should be. Have you managed to use a precision movement of the screw over a distance and then back so that a table contacts an indicator? This would tell you if possitional errors were building up or not.

One screw slipping on a multi roller set up is impossible if they are threaded, but not grooved. For any threaded roller to slip, "all" rollers have to slip. If they didn't, they would be out of time with each other and would bind. All lash would be removed to the point of nut lockup to the shaft. You cant adjust lash by moving threaded rollers front to back, just in or away from the main shaft. Moving the rollers forward and back would just make them roll to the new engagement point (carried over from another thread).

I'm thinking that if the roller carrier had one roller spring loaded to the screw, even if there were irregularities it'd ride through over them and maintain a steady preload. The rollers would never be loose enough to slip (hopefully).

Interesting thread, I just read all 134 posts and there are a lot of good observations.

I have to commend Posix on his enthusiasim and tenacity. In his rush to improve the state of the art he has overlooked something. As he says, visualizing how it works is difficult. The hardest to visualize is the contact path of the helix.

By the very nature of a screw thread, in order to advance a nut (of any type) there must be a thrust along the screw axis, that means that the flank of the screw thread must push against the flank of the nut thread. In an Acme thread that is the full face of the flank. In a roller screw the peak of that force is at the plane that runs through the axis of the two screws. That plane defines a line contact. That means there must be sliding contact between the flanks the threads.

When the mating thread is rolling (as in a roller screw) the contact is smaller ( a line versus a helical plane) so there is considerably less friction than say, An Acme screw. However there will be more friction than a ball screw.

When a ball screw advances a ball nut, the contact path along the flank of the ball screw is a point contact. Any plane tanget to a sphere results in a point. So the compression force between the flanks of a ballscrew/nut increases the force on the balls, causing the balls to rotate, resulting in less friction. Any intermittent contact between adjacent balls will not cause any measurable drag due to the low forces involved.


Having said all that, I think persuing the roller screw idea as a DIY project is great and I would like to help. I just want to make sure we know what the results will be - a leadscrew that is less efficient but has more load capacity than a ball screw.


Posix, If you want to forward me your design info, I will see if I can make up some prototypes for you in my shop. We are currenly running a batch of small leadscrews and I might be able to squeeze in some for you.

BTW, I have designed roller actuators and it took me weeks to get a grip on the whole helical contact path problem.

I, too, commend Posix for his tenacity. But I now wonder if he and all the contributors to this thread have been overtaken by a new development in the UK.
Marchant Dice is advertising on ebay various lengths/ sizes of "trapezoidal acme ballscrew leadscrews".
This is obviously a universal answer to all the problems - trapezoidal balls, anyone ? :rolleyes:

..erm....maybe the trapezoidal refers to the rod being bent into a trapezoid so you get instant x-y motion with just one leadscrew?

Not 100% sure. I think that accross the pond in England trapazoidal screws meen acme thread.

Here in England Trapazoidal Screws are indeed ACME Threads, we always like to make the words as long as possible ;->

Here in England Trapazoidal Screws are indeed ACME Threads, we always like to make the words as long as possible ;->

Probably because The Acme Pewter Tuning Fork Company had tied up the tradename.

Why do I have an eerie feeling that everyone hates me right now?
Are you kidding me? I think I might have more respect for you than anyone on the plannet right now! You had an idea, tried it, spent big money on it, failed, kept trying, spent money, kept trying. That's awsome! I have learned more from this thread than any other on this board!

Geof - is it one of these we need to tune up our servos ? :D

Hi everybody.... Well, I will try to help from other point of the world.

I don't Know if anyone asked it, the helix of the roller and the screw must be in different direction? ...if not, will it have friction problems?...:confused:

Try to put two screw together you will realised that move in an angle to get the tread fix correctly.


I was playing with CATIA And get this pictures..

Glossary: comment Spanish, (English)
right thread = clockwise
left thread = counter clockwise

also I made a formula to get the correct fix for the thread angle, considering the different diameters.

Geof - is it one of these we need to tune up our servos ? :D

Who was the cartoonist who created The Acme Pewter Tuning Fork Co? I am under the impression he was British (English)?

Posix;
There are many trite sayings about he who never dares never gains and all that stuff; mostly becasue in general they have a large element of truth. Sometimes after chasing an idea with lots of time and sometimes lots of money you realise not only is it not going anywhere it never could go anywhere and feel somewhat chagrined. However, if you never pursue anything you never get anywhere. I know on balance I am ahead with the return from ideas that did go places far outweighing what went into the ones that stalled. Probably most people who persisted are.

ggg: your roller thread is wrong way, both lead screw and rollers have normal, clockwise screw (right-hand?) threads.

geof thanks for the kind words of support but you did hit the nail on the head there by assuming I'm somewhat discouraged by this whole malarky.

It gets very VERY frustrating having to travel 10-20 miles to get something done on a lathe only to discover that the holes are 0.1mm out or somesuch when you need 0.01mm accuracy. Well, ok maybe 0.01 is too much to ask for but 0.05 is the biggest mistake this whole shebang will tolerate. More than that and it begins to wobble allover the place.

Yesterday I gave up completely and ask those guys to just make me 3 stupid acme rods and nuts so I can get this machine finished. Once I'm (hopefully) up and running I'll look into getting a 4th axis (somehow) and experiment myself. It'll be cheaper (at least no petrol costs and wasted materials at least) and quicker (cnc).

I, however, do believe my design works. I have built-in some adjustment to get rid of slip completely obviating need for teeny-weeny pinions but I need parts machined to exact specifications rather than 0.1mm out here and 0.1mm out there. 0.2mm is lots of slip, lots of play, lots of backlash, lots of waste above all, both money and time-wise.

I will be putting this on a back-burner for a while, at least untill I have a working machine to work with. But even THAT is proving difficult. I think these acme leadscrews will prove to be too much for my steppers, 20mm steel rod (their linear support can't take rods less than 20mm diameter!) 500mm long, weigh a ton. They did promise to make me an aluminium lead screw if this one can't be moved by my steppers and I suspect that will be the case...

ggg: your roller thread is wrong way, both lead screw and rollers have normal, clockwise screw (right-thread?) threads.


But if you have both normal , clockwise screw, you will have two contact points not axial aligne . The roller thread will try to get an angle to correct fix with the lead screw thread. is that a problem?
If you have different ways you roller an the lead screw forces work aligned so you will have less friction.

Originally Posted by posix
ggg: your roller thread is wrong way, both lead screw and rollers have normal, clockwise screw (right-thread?) threads.
----------------

If i'm understanding this correctly (perhaps i'm not but i think i am), having opposite threads, two equal sized rods whose threads were meshed would result in no linear movement as the driving rod would drive the other rod in one direction, but the driven one would "climb" the driver at the same rate it is being driven. (hope that made sense)

if we used this arrangement with satellite rollers that were smaller than the driving rod it would result in linear motion in the opposite direction of what we're used to. and if the satellites were larger it would go in the traditional direction, but with a speed REDUCTION. Also ggg's "backward" threads will will have forces axially along the rods in BOTH DIRECTIONS ALL THE TIME. Whereas traditional designs adjust out the backlash, this design FUNDAMENTALLY HAS NO BACKLASH.

So in theory, we could set our ratio of rod diameters to give us a very fine feed for a highly accurate machine.

A minor problem that i see with this arrangement would be the slippage between rods as always. I say "minor" because the greater friction would reduce slippage, and if our ratio is set to give us 1/10,000" feed and we moved 1/10,000" we'd have less slippage than that.

the major problem i see is finding someone to calculate the exact diameter of contact between the rollers so we'd know what our "gear ratio" was. And then the tolerances to which they would need to be machined to get anywhere close to that.

I hope i didn't post any wrong information, my head is starting to spin from thinking about how helixes interact :)

Hi Guys

What a thread! My eyes are burnt after reading all of these posts. I have been investigating the roller screw for a while now and have considered the following points which may be of interest to those considering making their own roller screw. The attached file was given to me by a friend with about 25 years on me. He knew I was into hobby CNC so he thought it may be of use to me. Sorry I can't quote the mag it is out of but the article was written by John Jardine. Thanks John wherever you are!

Using a roller screw which is as per attached file ie same thread used for rollers and main screw result in a lead multiplication of theoretically 2. Practically if you were to make this design you would use phosphour bronze rollers so as to minimise wear between the roller and main screw.

Using "All Thread" which is commonly available from hardware stores have relatively fine pitches (16mm, 2mm pitch for example) which when used with a simple nut severly restricts your rapid speeds. Having the roller design with a 2 x multiplication factor bring this back to 4mm (as a example using the 16mm, 2mm pitch thread) which gives reasonable rapid speeds.

"All Thread" is not well know for its pitch accuracy.

"Proper" roller screws (likewise for ball and leadscrews) are made from engineered steels made for high surface pressure use and are hardened with processes other than quenching when it goes cherry red. Most of us aren't going to be able to build a decent roller screw which will, performance wise, compare with a rolled ball screw.

You can buy precision rolled ballscrews from homeshopcnc very cheap (I'm not plugging them, probably other sources as well???) which have a lead accuracy of 0.1mm per 300mm which for hobby cnc is more than acceptable.

Go through this exercise and you will realise that the pain of researching this subject is for the fun of it alone (I did it, I had fun and learnt something new) and not a practical cost effective solution if you consider your own time and the finishing of your machine in a shorter time period valuable.

BTW my current machine uses 12mm x 1.75 all thread with 2 nuts in a wood housing separated by a spring. I'm not anti "All Thread" I just know that the next step is rolled ball screw and not making my own roller screw.

Cheers

Derek

I was really trying to get away from the bulky rectangular arrangement but that seems to be the only "diy"-able solution. Everything else requires a lathe. Not saying that this doesn't require one either but at least it looks a lot less forgiving!

I just found these intersting videos of the internal working of SKF roller and ball screws. I thought that others might like to take a look:
http://www.linearmotion.skf.com/en/0/cat-cat1829-Roller_screws.html
http://www.linearmotion.skf.com/en/0/cat-cat1830-Ball_screws.html
Under "related links" near the bottom of the page.
Check out the recirculating one. The rollers accually shift a thread each revolution!

Neatman

Thanks Neatman, cool videos. Don't forget to check out the PDF's as well. Heaps of interesting reading.

Cheers

Derek

Is there an explanation for the "jump"?

Is there an explanation for the "jump"?

:cool: if you look carefully in the "reciculating roller screw" the roller hasn't thread, only are "V" slots and the external thread have only 1 lead.

Looks like the "jump" corrects the relation between the roller an the external piece thread and the roller never goes out.

doesn't make sense. grooved rollers don't "go out" they don't have the gearing problem as threaded rollers. why would they complicate something needlessly?

It's so that the outer ring can be threaded instead of just grooved, which would allow gear ratios. There's probably other benefits, but yeah it does seem somewhat overly complex when it could be so much simpler. I'd imagine that design would have problems with high speed operation vibration and noise levels. Kinda neat how it works though.

doesn't make sense. grooved rollers don't "go out" they don't have the gearing problem as threaded rollers. why would they complicate something needlessly?

....could be a construction simplification. .... with the thread in the ring allow the ends nuts to be adjusted to avoid backslash.

remember also that with the tread in the ring the advance is higher.

The designs of the assorted commercial roller nuts discussed here can perhaps best be understood by considering them as what they are: planetary systems, directly analogous to planetary gearing.

The rollers are the planetary gears, the screw is the central pinion, and the nut/casing is the ring gear.

Sliding contact is inescapable since the three elements are all achieving the same pitch with different diameters, which means different "unrolled length" for the thread.

If we have a screw with a pitch diameter that has 1 inch of surface per revolution and planetaries that have .5 surface inches per revolution, and both elements have the same thread pitch, then in one full thread revolution there is .5 inch of sliding contact between the two members.

Sliding contact becomes much more extreme between the .5 surface inch planetaries and the outer nut, since the nut *must* be larger at pitch diameter than either the screw or the planetaries, but the thread pitch locks all three elements into rotating at the same "pitch speed" regardless how many inches of surface are involved in achieving that pitch length.

This is the reason why the higher grade, higher cost commercial units use not threaded rollers, but rollers with mating annular profiles. The planetaries become "decouplers" that unlock that thread pitch coupling.

The tiny gear teeth on the ends of the rollers (and the ring gears that drive them) serve no purpose other than to distribute the inescapable sliding contact so as to regularise wear and optimise part life.

To get full rolling contact with no component of sliding (apart from tiny sliding values as your two meshed arcs engage and separate), you would have to use rollers the same diameter as your screw and eliminate entirely the outer nut/casing.

Personally I'm rather amazed that no one has mentioned at all using rolled thread acme rod. The low 14.5 degree pressure angle would be a natural for this application, wasting far less input torque into deflection forces as compared to any 60 degree thread form.

The acme rod would also be far easier to support within the nut assembly since the flat crest would make possible use of independent plain rollers for support that could spin as fast as needed to match the surface inches of the screw major diameter.

I'm fiddling about with a CAD design for a compact acme unit, may post it later if there's interest and nothing interrupts my play time ;)


EDIT> had to correct the acme pressure angle. it's 14.5, not the full 29 included angle (duh)

Tiger

Go on, grandma, hit me with your best shot! :D

btw, that's grandpa :P

Barring snafus, renderings should be up in a few hours



Tiger

The Exlar stuff is quite expensive but you can get big forces and long life. Kerkmotion is low cost, low force. For similar sizes you are probably looking at a 10 to 1 or greater difference in cost between Exlar and Kerkmotion.

Personally I'm rather amazed that no one has mentioned at all using rolled thread acme rod. The low 14.5 degree pressure angle would be a natural for this application, wasting far less input torque into deflection forces as compared to any 60 degree thread form.

Tiger


Hey! Reread the thread, will ya?! ;)
Acme/Trapetz was one of the main discussions in the beginning of this thread. But I'm glad you're putting it back in focus. :)

Cheers,
Sven

btw, that's grandpa :P

Barring snafus, renderings should be up in a few hours



Tiger


few hours, few days, what's the difference, right?! (nuts)

Barring snafus...

The riders and codicils often determine ;)

Had a malf, lost all the work, starting over from scratch.



Tiger

Well, life got in the way and it took longer than it should have, but herewith the low grade renderings of the roller nut.

These are from a dimensioned dwg sized to fit under the table of my old bench mill. Had to reproduce (mostly from memory) an 8 yr old file that got irretrievably mangled when I tried to load the autocad 12 file into turbocad ;)

The idea is pretty clearcut: the hardened shoulder screws become off the shelf pinions for the rollers, the bearings are push (not press) fit in the bores and allow the screws to apply axial tension in opposite directions, taking up all backlash.

The rollers are from LaSalle ETD 150 which has a RC 32 hardness, 150k yield and machines very cleanly. They're annular, not threaded. The acme was 3/4" 8 tpi.

I know, the setup is not suitable for extreme loads or speeds but my goal was purely zero backlash in a design that was adjustable to maintain that condition. At the time I was doing nothing but specialty one-off and prototype work, the sort of thing you do slow, careful, and precise.

http://www.cnczone.com/gallery/data/500/medium/pinion.jpg
http://www.cnczone.com/gallery/data/500/medium/assembly.jpg

But this would call for very precise holes for those two bolts to be drilled in relation to the centre of rotation of the leadscrew. Very difficult to do DIY, wouldn't you say? Or are those bronze bearings eccentric off-the-shelf?

Also do you not need at least one more roller to keep everything paralell or are these two enough?

It was DIY for me. I already had the mill and lathe and was doing precision one off work :)

Didn't have room for the third roller, hence the odd design. The bushings pivot under load and the advance of the chord segment brings it into contact with the crest of the acme threads. Of course they pivot in opposite directions, so there is a total of four points of support for the acme. Passive self adjusting containment to keep the acme centered between the rollers.

This isn't really intended to be made with a hacksaw and an electric drill. Since so much else discussed here has been farmed out to local machine shops, this didn't seem particularly out of step with the thread.

The parts involve some tedious lathe and mill work, but there are only the housing, bearings and rollers that are complex at all, and should be well within the capabilities of anyone competent to be doing business as a machinist (imo, anyway).

I only had three things in mind with the design: low cost, zero backlash, and fitting it in the available space on my mill.



Tiger

ok ok show off grandpa!

so no biggie then. I'd only suggest either eccentric bushings or perhaps a slot cut for one of the rollers so it can be adjusted in case holes are misaligned or something?

one more thing, how are the bronze bushings when it comes to face-to-face contact between roller face and bushing face and bushing face and case face? this is where some would call for a thrust bearing since these surfaces are for transferring loads from leadscrew to the table. Would delrin do ok in this area? Or stick with bronze? finding a thrust bearing to fit in that space would be either a long search or an expensive one, pick your favourite...

That's the beauty of it. As long as the holes are parallel there's a few thousandths wiggle room on the spacing. All you lose if your hole spacing is off a bit is a couple or three thousandths of root/crest clearance or a bit of tooth depth engagement.

The housing is only 3 inches end to end. Anyone with a decent milling machine should be able to run 3 inch holes within at least .001 of true parallel I'd think.
The bearings are just boring lathe work (literally boring for the internals LOL). Cutting the radiused recess is just more lathe work using a faceplate fixture.

If I recall correctly what I was using for bronze bearings at the time was SAE 841 oil impregnated stock, sintered 80% compact. The stuff is in it's own right a medium duty bearing material, low friction and stable well beyond the needs of this application. The bronze pieces *are* thrust bearings, just not roller or ball; they're plain bearings :)

BTW, there's no reason this couldn't be used with more than two rollers. If I had room for a thicker unit, I'd step up to four rollers and stagger orientation around the circle.

Heh... methinks I see an excuse for more CAD playtime ;)


Tiger

I will try my idea with four rollers when i get Z converted... Posted it earlier here (http://www.cnczone.com/forums/showpost.php?p=130768&postcount=122) :) Could i expect 0 backlash if i keep the original leadscrew and go for "acme" rollers?

Not if you mount them like the ones in the drawings I posted. The problem with the threaded rod is there's nothing in the world to prevent if from just screwing down the pinion following it's own helix. It can only be held in one place by rigid support on both ends, unlike the annular groove version.

That's the theory as I understand it anyway, and the reason I opted for the additional hassle of making the rollers from scratch (which isn't really that much trouble, just tedious).

EDIT> hmmm. maybe I misunderstood your post. Just had a refresher look at your drawings and you illustrate annular rollers. If that's what you're using, then sure you could adjust alternate rollers to remove the backlash both directions.

Tiger

Reading through this thread I think many are over thinking this, or missing the point of rollers.

Goal: low friction, low to zero backlash nut that is easy and cheep to make.

Forget the round external housing, it GREATLY complicates making this, especially as a DIY, and is not required. Also forget having rolers with a helix, also not required.

The best solution I have seen is on the Hungarian site http://www.hobbycnc.hu/CNC/Otletek/Otletek.htm

The changes I would propose is using ACME threaded rod since it should be reasonable to get with good tolerances in Dia. and position along the length. This changes the form of the rolers to ACME of course. Just use an ACME single point insert instead on a lathe to make the form. Also single lead thread is fine, the rollers just have to be positioned pitch/(number of rollers) from each other.

Three or four rollers. As big as you like as long as they fit around the threaded rod without rubbing each other. The rollers will need to be preloaded in opposite directions aginst the threaded rod to eliminate backlash. The hungarian site did this by using allthred for the rollers. JBV's design could work, but he has to adjust 2 to match with spacers, the other 2 can be adjusted/tightened with the nuts on the other end. Actually thinking about it JBV's isn't that big of a deal. The threaded rod will flex to make contact with both shimmed rollers as long as the error isn't that bad.

EVERYTHING should roll. NO sliding or your lose most of the advantages. There should be bearings for the rollers, preferably angular contact because they will see thrust loads. JBV's design with bushings probably isn't that much improvment over a preloaded bronze nut for a lot more work.

JBV sorry for picking on your design, but it was a reacent post and a good referance.

Dale

That Hungarian site is nice; pure eye candy :)

I particularly liked the shot illustrating turning down the acme/trapetz rod to make a thread forming tap for the delrin.

You find some very creative stuff coming out of that area of the world. Some of the mechanical designs from when they were recovering from WWII are pretty mind blowing. They seem to have a lot of out of the box thinkers.



Tiger

The best solution I have seen is on the Hungarian site http://www.hobbycnc.hu/CNC/Otletek/Otletek.htm


Dale

Dale,

This web site is owned and operated by a web page pirate. I know personally some of the work and the photos he claims to be his own. Stating that fact, I'm not sure this guy knows what he's talking about, just stealing others work and presenting it as his own.

Good ideas? Maybe. Shameful plagiarisms, definitely!

Greg in California

I would happily change the link to another, especially if it was the original, and and in english :)

I was thinking this might be getting expensive when compared to the rolled thread ball screws.

So a quick cost check for a 36" screw:
Ballscrew from http://www.homeshopcnc.com/page2.html
$1.20 an inch = $43.20
2 nuts (or a single preloaded nut) $50 ($55)
Total $93.20

Roller screw (McMaster prices):
5/8-5 (matchig above)= $36.32
6) 5/16 ID bearings @ $5.33 = $31.98
5/16-24 threaded rod = $5.89
box of 5/16-24 Lock nuts = $4.84
Total $79.03 Plus the rollers, Material and machining

The roller screw can be more expensive or cheeper. Don't just assume that because you made it it will be cheaper. I think the tolerance on both are about the same as well.

If you can make the rollers yourself (or a friend will) and you can find a cheap supply of bearings it is probably a good alternative. You can also start with a cheap delrin block for a nut and upgrade to these later if you leave the room.

The other thing I like about the plain threaed rod or ACME rod is the easy solution for end supports. Just put it through a bearing and tighten nuts on both sides. Do this on both ends and you can pre tension the screw.

Dale

I would happily change the link to another, especially if it was the original, and and in english :)

Dale

Ok Dale,

Here's where it REALLY was dreamed up and designed... Dave Kush is the man who did the real ACME thread tap, the linear bearings, and the riser arms for the "Z" axis supports. I know, I've been in his garage several times and have seen the products.

http://www.buildyouridea.com/cnc/hblb/hblb.html

By the way, It's in English as Dave lives in California.

Greg

sorry but on that site I see not a single reference to roller screws. yes it is in english and in california. do you have something against people outside of california who don't speak english? because that's exactly where rollerscrews were invented, in germany where people speak german. :bs: if we followed that "logic" then CNCzone.com should be called PLAGIARISMzone.com, since this place is FULL of ideas and solutions from other places. but then that's the beauty of it all, we get to share our finds. can we refrain from bickering in this thread who did what and where and concentrate on roller screw designs alone please. thank you. (wedge)

dale on the issue of prices - you have factored in some bearings at $5.60 a pop. if you lose those and go for delrin or bronze bushings you save a big chunk (I'd say $5 instead of $31.98). also what is this: "5/8-5 (matchig above)= $36.32"? at first I thought it was the threaded rod but then you list that as : "5/16-24 threaded rod = $5.89".

If the item at $36.32 was a threaded rod of some sort I would love to see that! Must be made of gold and encrusted with diamonds! :D

sorry but on that site I see not a single reference to roller screws. yes it is in english and in california. do you have something against people outside of california who don't speak english? because that's exactly where rollerscrews were invented, in germany where people speak german.

But that site is in Hungary, and 4 of the photos deal with linear bearings and a tap made from an ACME threaded rod. That's all I was pointing out. I can post the photos I'm talking about if you like.

I know what this thread is about. I was just sorry that a web site was taking credit for work they didn't do.

First 2 photos are here: http://www.buildyouridea.com/cnc/IG-88/phase_eight/phase_eight.html

More are here. http://www.buildyouridea.com/cnc/IG-88/phase_five/phase_five.html

I only wanted to set the record straight. Not bickering.

GAWnCA: Understood. I can't referance that design very well without pointing to the only page I know of with it, or copying the images. Exactly what you don't like about the page. Point is a bit moot, I tried to edit my previous post to clarify for posix, and couldn't... Limited time to edit posts.

Posix Clarification
Roller screw (McMaster prices):
5/8-5 ACME rod= $36.32 (matchig size pitch and tolerance)
6) 5/16 ID bearings @ $5.33 = $31.98
5/16-24 threaded rod = $5.89 for mounting rollers
box of 5/16-24 Lock nuts = $4.84 for mounting rollers
Total $79.03 Plus the rollers, Material and machining

The threaded rod is "precision" rod with equivalant error specs (2C) as the rolled ball screw. I know you can get $5-10 ACME rod, but what size and pitch errors?

Yes you can probably do a lot better if you can find some surplus bearings. Regarding bushings, As I mentioned I think you lose most of the benafits of making a roller screw if you go back to using components that slide (bushings) as opposed to roll. What is the differance between the rod spinning in the nut, and the rollers spinning in the bushings?

I guess my main point is for the features I expect from a roller screw you are very close to the price of the rolled ball screw (same feature set). UNLESS you find cheap bearings and cut the rollers yourself. And I figure that any compromises in features isn't worth work and cost of moving up from a delrin block for a nut.

I'm not trying to argue aginst making roller screws. I just want to figure out the cost/benafits. It just seems like it is really easy to reinvent the wheel for $5 and 3 hours of time when you could have bought a ready made one for $7 and 2 hours of looking. But for may people the exercise/fun of making somethig yourself is worth a LOT more than any savings. Other people just want something that works. Plus you can make the roller screw to your needs, size length pitch.

Dale

^^^

Was me, wrong login and didn't realize it.
no wonder I couldn't edit my previous post.......

:withstupi

Well here are 2 more links with the roller design I favor... Both also from hobbycnc. :rolleyes: It appears to be post of various peoples router builds.

http://www.hobbycnc.hu/CNC/Mech8/Mechanika8.htm
http://www.hobbycnc.hu/CNC/Mech12/Mechanika12.htm



Dale

you are right as far as those costs are concerned. much saner to go with a rolled ballscrew and nut assembly. I suspect dale was sent here by kgb to destroy our morale. no, in fact I think he's a witch! get him!!!!!

back to roller screws.

all manufacturers use bushings for rollers to roll in. I guess it's all down to friction and friction is much lower on a small diameter shaft like that found on a roller (mine were 3-4mm diameter) since the surface is tiny really. of course thrust bearings would be ideal if you had some spare lying about.

p.s. my test roller screws cost me about $20-$30 apiece in the end.

That wiring track in the mechinika12 pics caught my eye. It may be some commercial product, but what it looks like is a piece of retired firehose that's had parallel ribs stitched in it. Nifty idea for economy track.

Always fascinates me to see out of the box applications for existing materials and products, and that idea certainly qualifies.



Tiger

you are right as far as those costs are concerned. much saner to go with a rolled ballscrew and nut assembly. I suspect dale was sent here by kgb to destroy our morale. no, in fact I think he's a witch! get him!!!!!
:p

all manufacturers use bushings for rollers to roll in.
LOL something I totally failed to notice! I guess mechanical advantage plays a lot larger roll than I guessed. I'd still hold out for bargan bearings though. ;)


That wiring track in the mechinika12 pics Looking at that gave me the idea of canabalizing one of those really wide Stanley Fatmax Tape Measures (http://www.homedepot.com) for making something like that. The tapes only bend 1 way, and will self support like shown. Put it and your wiring in a cloth? tube and your good to go.

Dale

An interesting read...

I am designing and building a device that must lift large loads with a minimum of torque, yet fit inside a particular envelope. Low cost is also important. Some calculations showed that a brass nut on ordinary threaded rod, 3/4"-10, required too much torque. Smaller rod, 1/2"-13, reduced the torque almost enough but the rod would not survive the required RPMs. Some redesign will allow the rod to be fixed at each end, with the nut turning in the center. Fortunately for me, backlash is not an issue.

To reduce the torque required to turn the nut, I started looking at roller screws, which eventually led me to this forum/thread.

I can identify 3 different types of 'roller screw'. There are two commercial types, the planetary type and the recirculating type. The third type is the DIY homebrew type, using freely rotating rollers fixed around the perimeter of the lead screw. I am not sure though that the homebrew type should be called a roller-screw, as it functions quite differently from the commercial types. The comercial types transfer thrust loads from the shaft to the nut through the rotating rollers. ALL the load is transfered through rolling contact. In contrast, the home-brew type modifies the nut to attempt to reduce friction by reducing the size of the sliding surface. All the thrust is transfered through whatever bearing or bushing is used at the end of the rollers, which is quite different from true roller screws.

Now, is there a way for the do-it-yourselfer to use ordinary threaded rod for a true roller screw? I suspect not.

Earlier today, I built a very simple planetary roller screw to examine how the various parts moved relative to each other. I used a short piece of 5/16"-18 reddi-rod, some 3/4" long 5/16"-18 set screws and a chunk of delrin for the 'nut'. Some quick measurements and some simple geometry said I needed to bore a 0.790" hole into the delrin, and then cut 18 tpi about 0.040 deep in the resulting ID.

The results were quite educational. These sizes allow for 6 rollers to be inserted, although this results in the 6 rollers just jamming together, forming an overly complicated nut. With only 5 rollers, things actually start to roll. The trouble is, things also start to shift axially. The shaft does roll nicely inside the nut, and it does advance one pitch per turn. The problem is, the rollers advance 1/2 a pitch in the opposite direction!

This leads to the subtleties hidden in the commercial designs. If the rollers are threaded, the nut and shaft must be multi-start, which precludes the use of ordinary threaded rod. If the rollers are not threaded (ie, annular grooves), the rollers must recirculate and the leads of the nut and shaft must match. The Steinmeyer design is an interesting variation that combines threads and annular grooves into the same roller so that the lead of the shaft and the nut (actually, no lead inside the nut) do not have to match.

Now, I have to decide if I copy the Steinmeyer design or whether I fabricate a recirculating roller type, as these are the only options if I wish to use ordinary threaded rod.

If the rollers are not threaded (ie, annular grooves), the rollers must recirculate and the leads of the nut and shaft must match.

You can have both the rollers and nut be annular grooves if you use multi-start rod.

Dale

You can have both the rollers and nut be annular grooves if you use multi-start rod.

Agreed. The number of rollers equals the number of starts.

So, did this come to a grinding halt? Things were going quite nicely and then....

Nothing...

Hmmm...

So, posix - What does your finial design look like?


you are right as far as those costs are concerned. much saner to go with a rolled ballscrew and nut assembly. I suspect dale was sent here by kgb to destroy our morale. no, in fact I think he's a witch! get him!!!!!

back to roller screws.

all manufacturers use bushings for rollers to roll in. I guess it's all down to friction and friction is much lower on a small diameter shaft like that found on a roller (mine were 3-4mm diameter) since the surface is tiny really. of course thrust bearings would be ideal if you had some spare lying about.

p.s. my test roller screws cost me about $20-$30 apiece in the end.

This the first time I've ever seen one. I wonder if and when the major machine tool manifacturers are going to use these?

Posix,

Your design looks so clean and simple. Do you think there is a need to make gears and outter threads? It seems to me that what you have in these photos would work just fine as long as ware can be adjusted for.


I haven't forgotten the planetary gear or the outer threaded shell but this was a test just to see how easy it turns and if the concept is ok.

The next phase is to take one of the cutaway drawings of a real roller screw together with all the elements and make the real mccoy, custom threaded rod, little threaded and toothed rollers, outer threaded shell, the works. Can't wait 'till thursday... :banana:

What about using dies from a thread rolling head.
They should have a long life span and should help to keep the form on the rod

Originally Posted by posix
ggg: your roller thread is wrong way, both lead screw and rollers have normal, clockwise screw (right-thread?) threads.
----------------

If i'm understanding this correctly (perhaps i'm not but i think i am), having opposite threads, two equal sized rods whose threads were meshed would result in no linear movement as the driving rod would drive the other rod in one direction, but the driven one would "climb" the driver at the same rate it is being driven. (hope that made sense)

if we used this arrangement with satellite rollers that were smaller than the driving rod it would result in linear motion in the opposite direction of what we're used to. and if the satellites were larger it would go in the traditional direction, but with a speed REDUCTION. Also ggg's "backward" threads will will have forces axially along the rods in BOTH DIRECTIONS ALL THE TIME. Whereas traditional designs adjust out the backlash, this design FUNDAMENTALLY HAS NO BACKLASH.

So in theory, we could set our ratio of rod diameters to give us a very fine feed for a highly accurate machine.

A minor problem that i see with this arrangement would be the slippage between rods as always. I say "minor" because the greater friction would reduce slippage, and if our ratio is set to give us 1/10,000" feed and we moved 1/10,000" we'd have less slippage than that.

the major problem i see is finding someone to calculate the exact diameter of contact between the rollers so we'd know what our "gear ratio" was. And then the tolerances to which they would need to be machined to get anywhere close to that.

I hope i didn't post any wrong information, my head is starting to spin from thinking about how helixes interact :)
If you revisit the European site you will notice the planetary threads are much larger than the rod ??? His system must work because he has it on his router. So the different diameters must come into play.

What a great idea :thumbsup:

Since my last post I have tested some yheories put forward on this thread.
The simplest form would be A rod and three rollers. All the same diameter.
If the rolls are different diameter to the rod you will have a problem with the helix angle that will cancel out straight rolling action. You would not need multiple starts. If the tree rolls were 120deg apart, one would have no axial movement to eliminate backlash in the assy , and the other two could float allowing them to locate properly once the rod is inserted.
Two rolls would be fixed radially and the third would be adjustable to compensate for wear. Lastly I removed the pionty end of the threads to ensure that there was no interference at the root of the threads
It is simple and it does work.

pictures? a picture says a thousand words you know...and it is a strictly enforced policy on this forum that all posts without pictures will be removed and posters banned. for ever. and outcast into the depths of sahara desert. with only a pot of peanut butter as survival kit. and a pack of hunting dogs sent after them to track them down and devour their remains. :D

Sorry, I don,t have a digital camera, but then with an imagination like that you should not need pictures.:wave:

wjfiles,

Do you have any drawings of your design? That would give a pretty good picture of what you have accomplished. It sounds great and a pretty simple way of making a lead screw with no backlash.


Sorry, I don,t have a digital camera, but then with an imagination like that you should not need pictures.:wave:

I don't have drawings, what I made was a copy of the setup on this link.http://www.hobbycnc.hu/CNC/Mech12/Mechanika12.htm. The 3 roller setup is small and simple
I also worked out that the rolls can be bigger than the thread but only in factors of the diameter, ie, 8mm rod and 16 or 24 mm rolls but 16mm rolls would have to be 2 start and 24mm would have to be 3 start.

I don't have drawings, what I made was a copy of the setup on this link.http://www.hobbycnc.hu/CNC/Mech12/Mechanika12.htm. The 3 roller setup is small and simple
I also worked out that the rolls can be bigger than the thread but only in factors of the diameter, ie, 8mm rod and 16 or 24 mm rolls but 16mm rolls would have to be 2 start and 24mm would have to be 3 start.

i don't think it can load high torque .

I've seen that site and that's why I was saying I don't see the need for plantary gears and all that stuff. Posix's closed system would be great where there might be a lot of metal chips getting in the threads. After seeing this thread I went out and took just 2 pieces of all thread and stuck them together, holding them in my hands and found that backlash would be zip, zero just one on one. So 3 outside rods at, like tou said 120 degrees, with one being adjustable agains the center rod would or should be rock solid. I wonder if it will work with ACME or Ball rod as well?


I don't have drawings, what I made was a copy of the setup on this link.http://www.hobbycnc.hu/CNC/Mech12/Mechanika12.htm. The 3 roller setup is small and simple
I also worked out that the rolls can be bigger than the thread but only in factors of the diameter, ie, 8mm rod and 16 or 24 mm rolls but 16mm rolls would have to be 2 start and 24mm would have to be 3 start.

i don't think it can load high torque .

Why not?

Why not?

because nut make by copper

So to avoid that just use the same material as the drive rod and you should be in good shape. I do't think I would have drilled through but turned down the rod and then threaded it. Use lock nuts on the ends and it not only hold the planetary rods but the but the bracket together as well. You could even fit bearings over the ends. Check out Posix's idea at http://www.cnczone.com/forums/showthread.php?t=13593&page=2
He's on the right track. Simple and yet very functional.



because nut make by copper

with reference to roll threads,
would anyone know the pitch circle diameter of a 12mm threaded rod

This is an example, sized for ¼-20 threaded rod. I am making a prototype with the ends of the rollers turned down to .125 and will take a photo. You may or may not get any thing out of looking at this.

Thanks for the drawings. We await your results.

Jason

wjfiles - according to my "Newnes Engineer's Reference Book" 1st Edition 1946

12mm x 1.75mm pitch.... 10.863mm
12mm x 1.5 mm pitch.... 11.026mm

Hope this helps.
John

This is an example, sized for ¼-20 threaded rod. I am making a prototype with the ends of the rollers turned down to .125 and will take a photo. You may or may not get any thing out of looking at this.

Hey Volts. I got a lot out of your drawing. Now I've got to learn how to draw in AutoCAD. If you could place that in a can type enclosure, you would have the same operation and yet the threads would be somewhat protected from flying chips and dust. Posix's idea was good and so is yours, simple. Let us knwo how well it works out.

GAWnCA,

Ouch! That hurt. Are you offering a class in AutoCAD, maybe I’ll sign up… Seriously I consider my self a CAD dabbler. If you got something out of the sketch I’m glad. If I get something functional put together I’ll do a more complete drawing and some photos.

Ouch? I didn't mean to ouch you. LOL I do need to learn CAD. I did understand what you drew up. All these types of ideas give me and others ideas to maybe start dabbling and improving. Who knows, we may never need to buy another $35.00 roller nut or spring loaded lead nut.

GAWnCA,

Ouch! That hurt. Are you offering a class in AutoCAD, maybe I’ll sign up… Seriously I consider my self a CAD dabbler. If you got something out of the sketch I’m glad. If I get something functional put together I’ll do a more complete drawing and some photos.

Wow, I just read that whole thread. I think it should be a requirement that anyone who wishes to post in this thread read the whole thing... I cannot count how many times people came to the conclusion that in the designs which involve a pitch on the rollers of this design that any slippage will cause the nut to become inaccurately located. I remember at the beginning of this whole thing I was wondering how anyone was going to make it work unless they cut the rollers with straight grooves.

Anyway, this has all been beaten to death... so I'll try to bring my observations to the table hopefully to further our development on this type of screw/nut homebrew design.

First and formost, I think this design should have at least 4 rollers, and if there are more they should be an even number. Maybe I'm being too picky but if you use only two rollers and preload them laterally you essentially are twisting the nut... (when looking from the top of it, I'll do a FEA model for you guys if you want to see what I'm gabbing about) what probably doesn't matter if the rollers are of any substantial length, such as 3 times the diamater of the leadscrew or more, but still, the forces aren't even on anything in the assembly. A member posted saying that putting preload downward on the threads removes backlash but I question how much preload is required to do that and how easy can it be maintained. With only two rollers I would think that the leadscrew would really want to deflect off of center to relieve its preload, which then would transfer into unwanted friction between the leadscrew and whatever is keeping it on center. Not effecient, but this all comes back to how accurate you want everything to stay, and how cheap you are.

With four rollers the leadscrew would always keep its center location, and the forces preloading the threads would be evenly distributed thus keeping the friction between those surfaces to a minimum. ACME threads have been mentioned a few times and I cannot help but think that would be a nice way to go about this design. I figure the rollers do not have to be all that big, but the allowable pitch of the leadscrew would have to be determined using the diameter of the leadscrew to make sure the rollers wouldn't be scuffing too much as the helical thread started deviating away from the straight cut profile on the rollers.

I think I may not be describing this well. The last couple posts were about CAD. My two cents, programs like ACAD are awesome for what they are good at but if you want to get into real fun you have move onto parametric solid modeling. I design for a living and I'm currently running SolidWorks. I'll never go back. Inventor by AutoDesk is also very similar.

I should model something up and throw it on this thread so I don't get stuck in the desert with peanut butter... mmm... peanut butter.

Its my first post guys... I love this site. I've been restraining from doing the typical noob "I wanna build a mill what should I do first" post and I've been researching this site and others (such as 5bears.com) for as much as I can learn before I start requesting anything!

So far this forum has posts from some really good minds out there. I look forward to reading more of your ideas people!

Dave

Hi Dave, good show old chap, having read the whole lot you at least know what direction our thinking is gravitating to.
One point you forget from the posts, if you try to mesh Acme screw thread with Acme form roller, the helix angles being of opposite hands at the mating faces will squeeze each other apart and only the tips will be in contact.

However if you use vee thread they will mesh full depth, allowing for tip rounding to prevent bottoming, and so will lend themselves to full screw thread face contact, as they rely 100% on friction of the flanks to cause rotation of the rollers.

There is a certain amount of friction, even though the thread and rolls are rotating in opposite directions while meshed, but this is of a low order compared to the sliding face contact of the Acme thread and a bronze nut.
The problem lies in getting precision Vee threaded rod as opposed to off the shelf readily available Acme transmission screws.

To use Acme threaded rod, for accuracy as opposed to transmission, would mean you would have to thin the flanks of the roller to allow penetration into the thread, which means the roller would have to be split to allow squeezing together for anti backlash application, see my sketch from an earlier post.
All of this becomes simple when you only want to design for frictionless transmission, as the vee screw and threaded roller will work very efficiently and simply together.

BTW, you only need 3 rolls to centralise the screw, 4 or more are irrelevant, but might become necessary when the rolls are made a much smaller diam.
Ian.

I have watched this tread with interest hoping to see a simple and easy to make linear motion assembly.
The best in my opinion is the first photo posted by POSIX.
Three rolls with a series of grooves the same form as the thread on the lead screw.
I believe that one of the "planetary" rolls would have to have NO axial movement. This would eliminate backlash.
The other two would need to float a distance at least equal to the pitch of the lead screw.This would make sure that all rolls constantly meshed correctly.

The Planetary rolls would need grooves with a form slightly modified to allow for the helix angle on the lead screw if you want a perfect match.

One roll would need to be adjustable radially to maintain a constant meshing of the threads.
I hope to make one over the next two weeks. I will use an M16 x 2.0
High tensile threaded rod and remove the points from the thread so that it cannot bottom.

WJF

Hi Wjf, sounds good, I would make all three rolls a close fit, end wise, and add thrust bearings at both of their end faces to prevent backlash.
The rolls will rotate to get position for all three to engage quite freely, and then they will each share the end thrust generatd by the screw.

Whatever thread you use, and for a trial run Allthread would be OK, the rollers only need to be a grooved vee form of the same thread angle and pitch.

Acme thread could be used, as it's a pretty accurate thread off the shelf, but the rollers would have to grooved to Acme form and the flanks thinned a bit to allow thread engagement.

With the flank thinning it means that the rolls would have to be split and tensioned endwise to grip the flanks of the thread, otherwise bottoming will occur, resulting in backlash.

I am looking at the possibility,( at his moment in time on paper), of making right hand Acme thread to mate with left hand Acme threaded rolls.
While they will rotate freely as the helix angles match perfectly, no motion will occur as the left/right hand mating will cancel out any forward movement of the nut, and so the screw will just rotate in the nut without going anywhere.

However if an internally threaded shroud is then made to mate with the OD of the rolls, we will have a sun and planet type of set up, whereby if the shroud is held stationary and the rolls allowed to rotate round the screw, it will cause the assembly to advance forward as the screw rotates.

The same would happen if the shroud were to be rotated, and the rolls were held stationary, the screw would advance or retreat also, similar to a ballrace and cage.

The necessary fits would add a lot of complication, when simplicity is really what is being aimed for.
I would be interested to see what progress you make.

At the end of the day, there will still be some friction as the faces of the rolls/screw will be a partial sliding and rolling contact, but not so much as a total sliding bronze nut/Acme thread set-up.
Ian.

I agree if it has more moving parts it will cost more and wear out faster.
I am sticking with the KISS method for motion.

Hi all, I just had a stroke of luck.
I am a great believer in "one picture is worth a 1000 words", and while scratching in the "too valuable to throw away" box I came across some goodies that I've been hoarding for years, knowing that someday they'll come in handy.

I have always been under the impression that Acme form will not mate with Acme form, when placed alongside one another, due to the helix angles being different, seems that was a wrong conclusion.

Long story short, I found a 300mm (1ft) length of 20mm diam acme left hand threaded screw rod, reclaimed from a dead vertical spindle molder, and a pair of short worm screws, from a worm drive gear boxabout, 40mm long and 20mm diam with right hand Acme thread forms.

So what? well the short worm screws are Acme thread form, the same size and pitch as the length of Acme thread, but different hand, how convenient.
Placing the rolls and the shaft together showed that the Acme left and rolls right hand threads mate perfectly for their full length, which is understandable as the helixes match on opposite sides.

However when the rolls, (both right hand threads) are placed together they also mated, almost to full depth, probably 90% engagement, which means if a roller is made to Acme form, but GROOVED not threaded, it will engage the threads 100%, good enough for the purposes of a very practical roller nut, without anymore than a pair of thrust races at each end to take the forward and back load.

For general transmission purposes, where accuracy of position is not an issue, the rolls can be made from the same off the shelf Acme form thread that they will mate with, but for high resolution requirements the rolls MUST be a straight grooved not threaded form.

As a picture is worth a 1000 words I'll add a few photos next post to illustrate the match of threads.
I think this is a major breakthrough as the Acme threaded off the shelf screw threads are much more accurate than the Allthread rod.
Ian.

In the interim of posix getting his thing done, I present unto you a roller screw design I've been working on, and plan on prototyping if I ever come up with a CNC frame I like (don't hold your breath). While it looks complex at first, I tried to design it from the aspect of as little complex machining as possible. The three "planets" around the central 1/2-13 UNC rod use a straight grooving instead of a thread/helix, so that there aren't positional errors due to minute slipping between the planets and central screw. Of course, you don't get the speed multiplication that threaded planets provide. The planets have inlaid bearings (cheap from VXB) which ride on 1/4" rod. I couldn't find inexpensive thrust bearings, so I figure take a pair of nice metric washers (the bore on an M6 washer is a perfect fit for 1/4 rod) and sandwich some small BB's, while fabricating a retaining ring out of thin sheet metal, ideally brass. Finally sandwich it all between some angle and it's good to go. If the holes in the angle supports are oversized, it should be possible to bring the planets into contact with the screw with an additional light preload.

How about Zach_G's design done on a drill press with a Thread "repair"file $13.99at ace hardware they usually come with 8 thread sizes ( thread Pitches: 11, 12, 13, 14, 16, 18, 20, 24 T.P.I. I say "thread size" as a thread repair file held at exactly 90 deg to the shaft makes thread sized grooves with no pitch. In normal practice they are held at the pitch on the thread that is being repaired. Chuck the "roller" in the drill press add a vertical guide rod up from the table to each side to steady and control depth. Set the press for a low speed.Then work it like Knurling, file in 6 or 7 grooves overlap by two or three grooves repeat down the roller then back up until you get to where you are flush to both vertical guide rods. holding the file by hand is actually a advantage if you get a good set of grooves to start as the file wants to follow the previous grooves.
http://www.acehardware.com/sm-general-and-reg-thread-repair-file-177-1--pi-1292851.html
I would imagine the same thing is available in metric.

Hi techno, I reckon if lathe facilities wern't available you would get by using the thread repair file, but at the end of the day you would be better off getting a lathe, any lathe, even a woodturning lathe with one of those X-Y tables for a DIY lashup.
Ian.

I got bored one day and made this. It turns rough (similar to an abec1) but then I made it using a 3 jaw without worrying about runout. Plus the bearing surfaces are the shoulder of a 1/4-20 screw and some flat washers. The flat washers are definitely very perpendicular to the screw axis either. Oh yeah, it also doesn't engage the 3/4-10 threaded rod because I seeemed to move the cross slide .11 for every groove, oh well... Despite all these issues it feels pretty smooth when loaded from only one direction and the total runout is about 0.005". For prototyping purposes it demonstrates how to make a cheap roller supported with angular contact bearings that could function decently for DIY if care was taken.

Hi Zach, I agree with you completely, the grooved rollers will work very well and accurately as a roller nut.
If it's only transmission you need, without accuracy, then the same thread can be used to make the rollers.

I have a collection of worms from worm drive gearboxes, which I used as a test set-up, and although they are an actual Acme thread they mate with one another very well, but for a roller nut where accuracy is required, the threaded form needs to be a grooved form as you have shown.

All you need now is to make a set-up accurately, running true with proper thrust bearings either side and you'll be home and dry.
The principle is pretty simple, and the beauty of it is it can be made by the very average shopworker with little or no skill.

The end result would be a screw drive with no backlash and no friction for very little outlay.
I'm contemplating this for a half nut set up for a lathe, running on the existing acme form leadscrew.
Ian.

hello,
this system look like a very good alternative
specialy for those who use overloaded servo (temperature increasing when holding the torque)
is there any one making a cheap kit with this system ? (16 mm and 20 mm screw)

if yes i buy it
because ball screw are VERY VERY expensive and very fragile too

Hi rokag, there is a lot of information on this type of screw drive.
Go back to the first post on this thread and read all 19 pages.

When you've got to this one you'll be as wise as we are without us having to repeat ourselves.

The method is quite easily made by the average shop worker with a bit of skill and some basic machinery, just depends on how accurate you want the end result to be.

The system will suit practically any screw thread size.
Ian.

Hi,Handlewanker,i did read them all,
but i am wondering if someone as done something between the Hungarian guy and SKF
, not a very sophisticated but some honest clean work for about 150 us$ and not 15 or 1500$ or more
something that can feet on a standard 20 mm and or 16 mm acme screw or ball screw
also did anyone have done something with trapezoidal screw (well if i can keep my screw will be nice after all)
actually i have no operational machine, you can do it yourself and you also can buy, i have no problem to rewind a big transformer and make my power supply, do you choose to make or buy your transformer ,your encoder, your servo, your driver, your g code generator etc....
here i touch the point
each one of us feel very comfortable for somethings and might think "but why to buy .. so easy to do "
In every job i was doing i was astonished "But why do they pay me for "
what you know is easy
meanwhile you might have the willing to understand what you are doing
about the trapezoidal someone mention that "it could be possible " but nothing else;
To make o good CNC we need no backslash
low friction because we have small motors
fast move because our soft are not hyper optimized
must be cheap because most of us get a limited "often by our wifes" budget
and the time as to be split between many entities
all of us we look about ball screw
So
is there any one making a cheap kit with this system ? (16 mm and 20 mm screw)
that is the question
thank
lucien

Hi Lucien, it will all dpend on your mechanical knowledge and ability to use a lathe mill etc.
Basically, all you need is an Acme screw and three grooved rollers with the same pitch but not being a thread.

Having read the previous post on this subject you will probably know that you can go two ways.
The first way is to use an Acme threaded rod and make three rollers from the same threaded material.

This will give you a "roller nut" that will drive with the Acme threaded rod, but will not be good for accuracy due to "slip" between the rolls and the threaded rod.
It can be used where a frictionless drive is required and accurate positioning is not important.

If you want a roller nut that gives accurate positioning and will have no backlash and is frictionless, then you will have to make three grooved rollers to suit the pitch of the threaded rod.
All of this has been explained in previous posts.

I don't know of anyone marketing a kit for the roller nut, it's actually easier to make than a bronze nut for the Acme thread.
I'll attach a rough sketch to show what the DIY nut looks like, if you want to have a go at making it yourself, or get somone with lathe skill to do it for you.
Ian.

Awesome Thread. Props to Posix and those driving this innovation.

Hi there.

I want to add my thoughts on this thread. In every diy roller nut design I've seen here the load force is transferred from ends of the roller pins. In commercial designs the load is transferred thru rollers to nut cover.

I think it might be possible to make such nut relatively easily with regular single start acme/threaded rod. With 3 grooved rollers a grooved nut with pitch of 2/3 of the leadscrew could be made. Of course a lathe with a proper boring tool is needed.

Rollers have only half of the grooves of the leadscrew.

Difficult to explain with my english but hope the picture will point out what I mean.

It has some merit, have you made one yet? Size is a consideration, cost is another. materials of construction etc. and of course how long will it last?

Hi Ian,
Your post, (224), interests me. I've just aquired a new 14-40 Taiwan lathe that I will be converting to CNC and if your design would give me less than a thou. backlash it would save about $350 for the Z axis.
Given an Acme screw diameter of about an inch, what roller major diameter would you suggest trying? Should I try for a roller form (shape) that exactly mates with the Acme screw?
Also I assume the eccentric adjusters would need to be on both ends of the roller shafts. Would that make for a difficult adjustment?
And, what material would you suggest for the rollers.

Thanks,
Ozzie

Hi there.

I want to add my thoughts on this thread. In every diy roller nut design I've seen here the load force is transferred from ends of the roller pins. In commercial designs the load is transferred thru rollers to nut cover.

I think it might be possible to make such nut relatively easily with regular single start acme/threaded rod. With 3 grooved rollers a grooved nut with pitch of 2/3 of the leadscrew could be made. Of course a lathe with a proper boring tool is needed.

Rollers have only half of the grooves of the leadscrew.

Difficult to explain with my english but hope the picture will point out what I mean.

This is exactly how the German design works. I was tempted to make one just to show you guys how smooth it rolls, but the time has been running away. I'll try to squeeze a free hour in somewhere...

Svenakela and Kuikkap,
It seems to me that this would require very exact machining of all the parts because there is no way of adjusting any of it.
In a commercial manufacturing plant that's probably not too difficult, but it's nearly impossible for the home machinist to hold the tolerances needed to accomplish zero backlash in the design you are suggesting.

I've read most of the posts in this thread, albeit quickly, and I'm not sure I understand why some are saying that four rollers would be best, others think we could get away with two, etc. It seems intuitive to me that three is the number for the home machinist, with one or three being adjustable as to distance from the center of the screw, that being determined by the carriage design and the method of anchoring it.
Seems to me that Ian's post #224 has it nailed perfectly for the home machinist; easy to make, fully adjustable, very low friction, and probably backlash below 0.001" !
As for cost, it's cheap; at least cheap for some types of conversion work. In my case I don't need to buy a new leadscrew. All I need to do is build the nut, (maybe $50-$60) and mount the screw in double angular contact bearings.
I still need to think about the diameter of the rollers and the profile of the contact with the screw.
Ozzie

Kuikkap: i like your design! I think it's much better than the one with small thrustbearings! What about splitting the rollers in 2 and adjust the backlash with a precise washer or a belleville spring?

I liked my earlier design too. But it wont work! Problem is actual linear travel per rev varies based on slide/roll ratio.

I tried to implement the 'recirculating' design by SKF based on the false assumption that the nut is grooved not threaded. But it took a while to figure it out how it really works. The nut is threaded with the same pitch than the leadscrew and rollers are grooved. Then why don't rollers roll out from the nut?
Because the nut RECIRCULATES. :) Thread in the nut isn't continuous, instead it has a groove parallel to screw. That groove is the place where recirculation happens. Every roller going thru that groove goes from thread end back to beginning.

So what this means? I think it could be fairly easy to build a nut for example metric threaded rod. M10 has 1.5mm pitch and M22x1.5 taps are easily availeable. Also regular M22x1.5 and M24x1.5 nuts are easy to find.

Rollers have to be custom made. I think this design needs at least 4 because 1 roller can be in the groove so 3 are needed to center the screw. Some sort of guides are also needed to prevent friction between rollers and to lift roller from groove to thread start (in SKF design end caps handle this).

Not a problem, I've already done a roller just as the German one on my manual lathe.
Kuikkap's version will work nicely, but you can skip at least half of the grooves in the nut and still have capability of tremendeous workloads . If you have a CNC lathe though, just go for it!

Not a problem, I've already done a roller just as the German one on my manual lathe.
Kuikkap's version will work nicely, but you can skip at least half of the grooves in the nut and still have capability of tremendeous workloads . If you have a CNC lathe though, just go for it!

Did you notice what is said of this german design?
http://www.steinmeyer.com/english/kataloge/pdf/drs.pdf:

"The basic principle, as developed at the German Aerospace Center
(DLR) (http://www.dlr.de), features a threaded spindle with a set of
rollers and a nut. Rollers transmit thrust from spindle to nut through
rolling contact, thus minimizing friction and wear. Rollers orbit at
constant speed while in uninterrupted contact with nut and spindle.
A patented feature allows combinations of fine pitch and coarse pitch
independently, and even combinations of left hand and right hand
threads in the same unit. Therefore, the feed rate per revolution is no
longer identical to the lead of the threaded spindle."

Think if rollers get stuck feed rate = pitch.

later on (about their own dlr derived system):
"The rollers work under "rolling friction" to the screw and to the nut. This fact causes a certain lost motion. Therefore there is no repeatability of
the feed. In case of repeatability is required, we are in need of an addi-
tional measuring system! On most applications, two end switches will
solve that problem."

Recirculating design does not have this problem.

Hi Kuikkap, I think for most people with home workshops and basic turning skills, the three grooved rollers for roll nut transmission on acme threaded rods, with one roller eccentrically adjustable and the rollers having thrust races to transmit the push and pull would be the easiest to make, ( you can also use vee threaded rods if acme isn't available, only accuracy over distance may not be so good due to vee threaded rod being produced with die heads).

I have a design in mind for my lathe half nut to replace the cast iron halfnut, (yes it's only got one half nut that presses down on the leadscrew), on my colchester Bantam lathe.

This design is straight forward and will have four grooved rollers arranged to pull apart like a regular half nut, with two rollers in each half nut and so will be able to transmit the thrust of the screw in a rolling motion as opposed to a sliding bronze on steel type.
The rollers will be grooved and not threaded so that accuracy of the pitch will be maintained as the rolls rotate.

The design of a home built rollnut has been well documented on this forum, and there are a few examples of actual roll nuts made that prove the principle.

At the same time I need to replace the crossfed nut as it's practically down to the gums and has about 3/4 turn of backlash, which means the threads are just about paper thin.
I'm toying with the idea of making a roll nut to replace the bronze nut, with four rolls about half the size of the crossfeed screw to keep the profile down.

This will practically last forever as the thread rolls as opposed to sliding, early days yet.
At least the back lash will be non existent, which will be a boon when withdrawing the crosslide rapidly in screwcutting, or for preventing "pull in" while parting off.

In actual fact the roll nut is easier to make than it's equivalent bronze nut counterpart.
The pressure on a single point screwcutting tool for internal screwcutting especially the Acme thread, is quite large and the tool is quite slender which makes deflection a problem.
Ian.

Hello Ian,
What you're proposing interests me.
Can you tell me more about the sizes you're talking about, what size lathe, what size cross feed screw. My 14-40 seems to hardly have space for a 5/8" ballscrew nut, at least an anti-backlash double nut, so I'm interested in your plan.

One thing I question is that you and others have talked about these rolling nuts having no friction and thereby lasting "forever". The fact is that they do have friction because they roll in a groove. A spot on the perimeter of the roller moves from the major diameter of the screw, rolling AND sliding into the minor diameter and then back out to the major. With all this happening in several grooves on several rollers, and there being the friction of the roller on its shaft, and the friction of the ends of the rollers in their housing, and all being adjusted to zero backlash, how frictionless is the scheme?

Regards,
Ozzie

Hi Ozzie, forgot you asked for the info on the proposed roll nut set-up.

While this is only a proposal, the parameters are met for "frictionless" driving capability.

The idler rollers meet the screw at the apex point of the screw helix and only contact the screw briefly as they sweep in and then out of contact, whereas the bronze nut is in contact for the total number of the threads and 360 degrees.

There should be at least 6 threads in contact with the rollers, which are not threaded but grooved, and to the same thread profile as the screw, eg 29deg as per Acme.

If precision 60 deg vee thread is available it would work better than the narrow grooves of the Acme thread and would lead to better thread depth engagement.

This is very important as the drive is by flank contact and so the rollers act as the flanks of a bronze nut, but are rotating instead of stationary sliding.

Due to the fact that the screw thread is a helix, which means that the thread flank will be half the pitch distance for 180 deg of rotation, so the rollers must have the end thrust faces of one pair of rollers half the pitch longer when they are situated either side of the screw, or be fitted with spacers if the rollers are made identical lengths.

This is easier drawn than described, but if you placed two rollers either side of the screw you will see that the roller ends are staggered by half the pitch.

In order that the screw will stay between the rollers it is necessary to have at least three rollers in contact so that the forces are centralised.

However this means that the top roller creates a high profile for the nut and would not fit in a confined space like a crossfeed screw of a lathe.

To overcome this problem I think that if you had four rollers smaller in diameter than the screw and paired on either side of the screw, this would still centralise the screw and have a low profile, but probably at the expense of going a bit wide sideways.

I would even consider machining the crossfeed screw slot wider to cater for this provided that it didn't weaken the casting vees.

It would be a perfect solution if the saddle of the lathe was designed with a wide crossfeed slot at the start to cater for a roller nut set-up, such as in some of the alluminium DIY CNC lathes being built, but in their case the slides are usually linear slides with lots of room between them anyway.

I would also seriously consider casehardening the screw as the rolling action of hardened rollers would probably cut into the flanks of a soft screw, even if made from tough alloy steel.

The rollers MUST have thrust ball bearings at their ends as all the force of the screw is directed to the end of the rollers and their supporting housing.

This design concept could be considered as state of the art for DIY backlashless screw drive, and would be a simpler more economical set-up than a commercial ball screw using DIY item capability, whereas the ball screw can never be made accurately, if at all, in the home workshop in a million years.

The other useage I envisage is for the lathe screw cutting half nuts, and in this mode
I think possibly only two rollers would be required with a couple of guide ball races riding on the top of the leadscrew threads to keep the screw centralised when engaged with the rollers.

The ideal design for this set-up would be in the centre of the bed between the slideing ways to ensure no side dragging force as in conventional lathe set-up.

There are many DIY lathe makers willing to "grasp the nettle" and produce new concepts for lathes, whereas the production lathe makers still follow the basic designs that practically date back to the beginning of the 20th century with only a few material changes to show progress.
Ian.

Hi Ian,
After agonizing a long time I've decided that all I have room for in my lathe cross axis is two 5/8" ball screw nuts, end to end in a common threaded carrier. I'll grind the corners off one and that will allow it to be turned enough to accomplish a near zero backlash.
I've used that setup before and it has worked fine.

However,
I will be putting together an Industrial Hobbies mill and I think I will use your ideas for the X axis. I have ground ball screws and nuts for the other two axis' but to buy similar for the X is out of the question for my budget. I have the brand new Acme screw so all I'll have in the experiment will be metal and bearing costs.
The roller nut needs to be a fairly precisely made item.
The distance between nodules of the rollers must be accurate
The relative end offset of each of the three must be accurate or adjustable
The rollers must be perfectly parallel to the leadscrew or adjustable

Adjustable sounds fine and might be great on a router setup where you might have good access to the nuts, but in the case of a milling machine access is usually difficult so highly accurate machining might be the only answer.

Ozzie

Hello all that might be interested in this thread:
My previous post was 3 days ago and I've thought about this a lot.
Last night I dreamed or semi-dreamed about it, half asleep, half awake.
I don't think a home shop machinist can produce a roller nut for an Acme screw that can achieve near zero backlash consistently. By near zero, I mean in the area of .0005", with the solidity needed for metal machining. There is simply too much going on there, too many factors, too many places for error. For the screw to push on three rollers equally would require incredibly accurate machining.
I've seen a lot of ideas in this thread but have yet to hear from someone that has made one, except for some pictures from abroad that appear to be installations on router machines.
My present thought is that if I'm going to use an Acme screw, I'll go the casting route; casting nuts with a graphite laden epoxy, (poor man's Moglice), and maybe even spring loading two of them to achieve near zero backlash.
I respectfully challenge any home machinist to actually make a roller nut for an Acme screw that has low friction, good longevity, and near zero backlash. You can draw it, but let me see you make one!
Respectfully,
Ozzie

You can!
I did in my bench lathe. Ok, it's a good lathe, but anyway...

Yes Svenakela,
I saw where you posted that before; over a month ago I think. But I've yet to see a picture of this nut or any specifications or any information whatsoever about how you made it.
It's a good lathe, commercial CNC??? Oh no, ,,,you said manual.
What is the backlash of the nut?
What did you do about axial and lateral offset of the rollers?
Tell us about the thrust bearings; better yet show us pictures.
Are the roller axles adjustable or fixed?
Did you use roller bearings on the axles?
What method did you use to space the nodules of the rollers?
How did you form them; a form tool; can you show us the tool you used?
Please supply some details to change my mind.
Please SHOW us!!!

Ozzie

I did one before to try it out, I think I have to go back to the workshop and make another one - and take a picture this time...
No thrust bearings and tight from the beginning. I've been saying "german design" over and over again. Take a look at it and give it a try, it's not as hard as it looks. :)

The other useage I envisage is for the lathe screw cutting half nuts, and in this mode
I think possibly only two rollers would be required with a couple of guide ball races riding on the top of the leadscrew threads to keep the screw centralised when engaged with the rollers.

Ian - I hesitate to ask, but what if you had your two "guide ball races" and only one roller ?
I'm thinking only of a small, lightweight machine, but what of it, in principle ?

Regards
John

Hi, so you see, you'll never never know 'till you give it a go.....Ian.

Hi John, yep I reckon that a single roller would work just as well.

I originally envisaged the pair of rollers I advocated like the normal half nuts coming together to share the load, but really they are working differently to the normal bronze half nuts.

As I see it using the pair of rollers, and in order to ensure that the leadscrew does not push out to the side, when the rollers close and are in contact with the screw, a pair of ball races contact the top of the screw, (would only works on the flat top of an Acme screw), and at 45 deg either side of the vertical, with the same arrangement on the bottom roller.

This will ensure the "half nuts(rollers)" stay engaged.

However having a single roller as you proposed would mean having something on the other side of the screw to maintain stability and ensure roller engagement.

This would probably mean a pair of ball races underneath the leadscrew pressing on the top of the thread, while the roller is pressed into the thread at the top, something like a three point steady.

I think this set-up would be very economical of material and simpler to make, while at the same time if there were at least 6 threads engaged with the roller and two ball races underneath it would work.

It must be understood that the whole thrust of the leadscrew is against the flanks of the roller grooves, which will rotate freely, and so the end faces of the roller must be lightly preloaded with thrust ball races to prevent end float.

The thrust ball races need only be in the nature of a bronze washer with holes to form a cage for the balls and two HARDENED, (no compromise here), flat washers to form the thrust bearing.

Whichever method is used, there must be a pair of forces coming together to hold the leadscrew, either with two rollers and two ball races either side, or just one roller and a pair of bearings on the other side, with the single roller and the ball races moving together like half nuts.

I think it is also possible to further simplify the arrangement by having the bottom ball races fixed and just a couple of thou clear of the leadscrew to allow clearance while traversing the saddle with the "half nut" disengaged.

In this design when the rollers engage with the leadscrew the ballraces will automatically be positioned on top of the leadscrew threads as the saddle traverses down the bed.

In the engaged position the leadscrew will be deflected downwards very minutely, (a few thou'), and be supported by the ball races and will not adversely affect the working of the lathe.

In the Colchester Bantam lathe I have of circa 1930 vintage, there is only one cast iron half nut that presses down on the leadscrew, with the leadscrew supported by passing through a hole in the back of the apron casting.
This was a bad design and done only for cheapness.

During engagement the leadscrew just rubs on the bottom of the hole so preventing it from being deflected downwards , not the most ideal arrangement.

Having a single roller would also simplify the making of the set-up, as with two rollers the grooves would have to be staggered half the pitch to allow for the helix angle of the leadscrew. This is not a problem with a single roller.
Ian.

This is an interesting thread- years ago a friend gave me a bundel of acme flat screws, and no nuts mmmmm. soo I made 3 inch carllors to fit the screws. I had a lot of old drill bits, so i fitted the right drill to the thread size and drilled the carllors and incerted the drill bit ends in the carllors and meshed to the thread and great fit. the tolerance was close to .002 and the back lash was about the same , some i used 2 carllors and a valve spring between them (preload ) and that worked better. good old fasion home made screws work great.

Hi 'coach, a friend of mine I used to work for had a similar problem when a bronze nut on a welding machine stripped, and I encouraged him to drill a series of holes along the side of the nut at the same pitch as the screw, (5/8" Acme 8tpi I think), and insert brass pins and a keeper plate to hold them in.

This worked so well that when the pins wore away it was the work of a moment to insert new ones and be back in business within five minutes.

For our next trick we're going to have a look at making a roller nut, with a single roller instead of the bronze nut.
Ian.

@handlewanker- great idea , if you use steppers the slop is imortant but useing opto encoders thats not always the answer . I set the optos up in the tool post rather than the screw. doubles as a dro and the reader

wow!!

ive just read all this thread and clicked all the links and read those, and ive just had the "is that the time i couldnt have been reading all this for 5 hours straght"

bt its solved my problem of building a cheap simple design. i may build a mockup this weekend ot see if it will work like i imagine.

thanks for saving me loads of cash (hopefully)

I just read all this thread and it is great. There are some very innovative ideas here. I recall seeing a "threadless" linear motion device advertised (http://www.amacoil.com/html/page3b.htm), but it used a smooth steel rod and had several ball bearings mounted at a slight angle to the rod axis, with the rod through the inside of the bearings. The bearings had a specially ground inner surface (probably spherical) and as the rod rotated, the bearings traveled down the rod length. The pitch was even adjustable by changing the angle of the bearings, and motion could be reversed without changing the rotation of the rod. It seems this is a threadless version of the same idea. I suspect it has the same problems with slippage of course, and may not be suitable for high loading.

hello,
someone as made it for a lathe x axis?
please send a picture
i will post one soon
but i have not test it yet

Hi Roka, haven't made one, just hypothesising a bit to get the ideal design.

The concept is quite straight forward, and a number of people have made them to some degree.

To work without slip and backlash requires grooved rollers closely mated with the threaded shaft, threaded rollers won't do as they will give false positioning due to slippage.

The principle has been documented on these posts some time back.
Ian.

Faaaascinating thread, and huge forum. Just wanted to put $.02 in here. Truthfully, I've only gotten through part of this thread. I'd played with this idea some time ago, the differential style, that is. Here's a rendering of the inside out version, intended to run inside a threaded bore. Of course, the rollers have grooves or threads, as has been covered so well here. The inner setscrew pushes on the center roller's Woodruff key to get your preload, one roller against the other two.
Several variants come up in a patent search.

I made a little demo with three bearings with 1/4-32 female threads inside, to run on a 10-32 threaded shaft. It worked remarkably well, and had an effective 100 or so threads per inch!
It was one of my favorite paperweights, stand it on end, get it turning, and it would spin down to the bottom, spinning madly, taking longer than sense would suggest.

Occasionally, it will be sticky, and I thought at first that it was crap in the threads. It was in fact the one roller jamming against the other. I think this may have been covered earlier in this thread, but the necessity for synching the rollers is important, especially on the differentials.

The drawing attached would require something like oldham couplings between inner races. I've been meaning to make one, but that's back burner #27.

Hi Mike, as you have drawn it and described, would this assembly be running in a threaded bore?

I hate to think of a bore length of a metre or so.

One of the problems in any alternative design is, will it compare or exceed expectations as far as useage or manufacture are concerned, of an existing design?

The ball screw is just about the simplest backlashless drive to date, and very cost effective too.

Lets have some more info on this design.
Ian.

Hello Ian,
The illustrated roller screw was intended to only move 25mm, and yes, inside a threaded bore. The long travel version I talked about, I could not find in my toybox today so I could take a pic, and I could not find the solid models I made (or it was made from sketches). I'll come back with it when I find it. Essentially, rather than having external rollers (convex on convex), the grooves/threads are inside a standard bearing that surrounds the (acme or whatever) screw. It is (now that I look it up, ha) almost indistinguishable from the Uhing patent US3789678.

I use a lot of precision preloaded ballscrews, though I've only done one system with driven screw, and hundreds with driven nut. I love them ;-)
There can be an efficiency increase with a roller screw, but that's not why I got interested.
There's a few (OK, darn few) situations I've come across, not exactly CNC machining, that require a very fine pitch. If you've ever worked with ballscrews of 1-2mm pitch, they're very finicky and difficult to get made right, and vary a lot in preload related torque, and are just so susceptable to contamination. You'd typically want this fine pitch to eliminate the need for a gear reducer, or to get a motor up into a higher rpm for scanning applications, where velocity ripple must be at a minimum.

I got caught up in the differential screws not so much from a practical standpoint, since it would be a very small market indeed that needs extreme fine pitch, but mostly as a fun excercize. Imagine having a 3/4-12 screw and achieving .02" pitch out of it! I don't know, it is just wild. One of my patents had this differential thinking, but it was a rotary traction drive.

Mike

The Rolscrew is designed for high load applications- obvious from the large bearing areas it uses. The Rolscrew could probably take a really heavy hammer blow on the carriage with no damage. A ball screw would immediately develop flat spots from the same hit.

The Kerk self-adjusting antibacklash nut is neat. One of those "why didn't I think of that" ones. Hope they patented it.

hello,

here the picture for my x axis lathe
the screw is a trapezoidale 20 mm x 4mm/turn
backslash none !!!
it turn with the finger
each axe is on excentric from both side
dimension of the roller 38 mm diam 52mm length
the shape of the ring for the roller as been modified it is not a perfect print of the trapezoidale it's a bit rounded
the work as been done by my friend giorgos lolos in athens
i have no doubt that it can handle the chariot of my lathe at a crazy speed
so it work
the electronic card you see is the new UHU HighPower servo controller diy for my spindle motor 3HP DC 180v 8Amp cont. and this work too !!

hello,
here the picture for my x axis lathe


Ha ha! Beautiful! Got to love a DIY rollerscrew. Niiice.

that looks great ..... could you post some construction picks. more details!!!!!! :)

hello,
i have done the drawing with solidwork but giorgos as to change some values and the system to prevent the bearing to block as been modified so i have to dismantle it to make the new drawing and i want to test it before so it will take a while
i will try to post something meanwhile

Hi Rokag3, that's exactly as I see it.

The one I worked on (just drawn ,rough sketch) had two rollers floating end wise and only one roller with ball thrust washers on each end face.

This allowed the one roller to be eccentrically adjusted on one end of the spindle, and the other end of the spindle used to put preload on the end face of the roller.

The other two rollers are allowed to float freely endwise so that they can accurately engage the screw thread once the eccentric roller is adjusted.

I think if the other two rollers do not float they could cause a bit of end wise misalignment if the pitch adjustment is not perfect, due to it being difficult to have the three faces of the rollers against the frame exactly.

This means that the leadscrew drive thrust is actualy being taken on the flanks of one roller and the other two just maintain centralised contact with the thread.

If it is desired that all three rollers take the thrust, then I would make an adjustable end bearing for each roller that could be adjusted to make the rollers engage the thread and end frame face simultaneously.
Ian.

Rokag3,

Does this design maintain position repeatability without loss?

After reading most of the posts on this thread it seems like it turned into a debate.

Nice looking work!

Jeff Alessi

hello,
it's nothing but a screw so if your rolls are parallele no problems the fact that they rolls avoid to use energy motion in heat so you can go faster and the lack of energy preserve your roller shape also and this is very interesting you can use your cnc lathe by hand while with a ball screw it's impossible and this will prevent your servo to get hot .
i choose to use 11 teeth so it work like a mathematical smooth function if you have room longer is better since i have not fixed it on my lathe i cannot make precise mesurement but with no load i cannot see any backslash with my comparator (1/100 mm) i can make it moving back and forth in 1 gradation i could not detect any backslah even with a magnifier so do not be affraid to make one
I will post a video soon

rokag3
Was it necessary to have all three rollers on eccentrics ?
While it might work with only one adjustable, is there an advantage in having all three adjustable, now that you see it working in front of you ?

John

if you can change the position of your axe there is no need to put 3 excentric !!!

rokag3,

Nice work!

What kind of bearings did you use with rollers? Are they inside of the rollers?
And what material did you use for the rollers? Brass, bronze or steel?
The metal seems a little yellow.

here pictures of the inside

it's much more easy to tune the rollers than realign the axe and the motor so i think that 3 excentric is a must

Motion Engineers,

After reading almost 300 posts about the Roller Screw All I want to know is if the simple design that is being suggested by rokag3 can maintain position repeatability without motion loss?

From what has been stated in past posts it seems slippage keeps plaguing this design.

What keeps that from occurring with the rokag3 version.

Please elaborate with simple a yes or no so that we don't have to read 300 more posts to find out it wont work.

How about a video with a dial indicator testing the design for five or ten minutes?

Thank You

In case anyone is still confused, the short answer is using 3 straight grooved rollers there is NO lost motion, (zero backlash), and very little friction, provided the rollers have thrust bearings on their end faces (both ends) where all the push and pull ocurs against the frame or housing.

I would go on to further state that it is only necessary to have one roller with end thrust races as the other two rollers can float end ways provided they are all adjusted to maintain contact with the screw thread, and this can be had by having only one roller with an eccentric adjustment.

If you apply logic to the design you will realise that if the three rollers were to be seized up and not rotate then the screw thread would rotate in the rollers like a bronze nut, but with rubbing friction, but as the rollers are grooved and rotating with the screw there is practically zero friction due to the rollers rotating.
Ian.

This design has it's drawbacks. Position repeatability is all about the zero clearance axial bearings in the rollers and sufficient preload of the rollers to the screw. Maintaining preload (without a spring system which would sacrifice stiffness) requires a precision ground screw, with hardened and ground rolls. All commercially available roller screws are very expensive for a reason. there's a lot going on in this design, just in the sense of appropriate materials.

I would not do it this way from a cost and life standpoint, in an automated high precision, high duty system, but for a DIY machine where you can adjust out the backlash occasionally, it is great!

I would not do it this way from a cost and life standpoint, in an automated high precision, high duty system, but for a DIY machine where you can adjust out the backlash occasionally, it is great!

i choose bronze so very tender metal big rolls and long one the profile will be optimised with the time the pressure for each contact is divided by the surface of contact on this design the 3 rollers work in both direction since the thread is engage by force with the excentric and both side are in contact (that is while we do round the profile) each roll as the same length but the thread start at 1/3 rd so r1=1/3-2/3/,r2=2/3-1/3,r3=3/3-0 so r1 and r2 are identical but mount reverse.
high precision yes
automatised high duty NO
but for hobbyst or prototyping workshop YES!!!

Ian - how do you feel about pushing the logic to having one grooved roller, fitted with pre-loaded bearings, one plain roller with eccentric adjustment and plain bearings, and one roller with plain bearings ?
Do you think this is two steps too far for a diy set-up, or perhaps not ?

John

EDIT ..but make the rollers x3 longer ?

Yet another design of a roller screw comes to mind.

What if rollers are made of same regular threaded rod than the lead screw.

To prevent slippage between rollers and the screw one or more groove(s) are machined axially to screw. To rollers are welded a little dot(s) that are ground as teeth to fit the groove(s) in the leadscrew. This loosely resembles a planetary gear without annulus but produces additional linear movement.

Yet another design of a roller screw comes to mind.

What if rollers are made of same regular threaded rod than the lead screw.


Yes, this works, but you must do your planetary gearing approach for the ratio to be right, otherwise the TPI will be all over the place. Larger rollers of the same thread pitch will get you a differential that makes for finer effective screw pitch.

...so you could skip the planetary synchronizer stuff if you have a brush dc motor and linear encoder feedback. Then, little changes in TPI as it runs mean nothing.

Yes, this works, but you must do your planetary gearing approach for the ratio to be right, otherwise the TPI will be all over the place. Larger rollers of the same thread pitch will get you a differential that makes for finer effective screw pitch.

Thats why planetary rollers are to be made from same rod as the leadscrew.
That will give sun (leadscrew) to planetary (roller) ratio exactly 1.

I think one groove and thus one teeth per roller would be enough.
In theory there would be a little slippage that would reset 3 times per round
if 3 rollers are used. Actual travel is exactly 2x pitch.

Agreed that it is 2x lead.
My point with respect to the ratio wandering has to do with the fact that there is a line contact on the thread faces, and the actual radius of contact has to be considered indeterminate along that line, which makes for uncertainty as to the ratio of roller to screw.
Your single tooth should definitely do the trick to get it back if it did wander some slight amount. Nice!

Hi Kuikap, I agree with your design proposal, up to a point, that is if you were to use threaded rollers using the same thread as the leadscrew, then your method of having the leadscrew grooved with longtitudenal grooves, like a continuous gear wheel, and the rollers having a grooved ring of smaller diameter on the end would definately cure the variation in drive position due to slip.

As a feasibility study, the leadscrew would be an Acme thread, off the shelf as it's available and pretty accurately made for most purposes, but it would then have to have the threaded part longtitudenally grooved with a gear form to mesh loosely with the rollers that are also made from the same Acme thread and grooved with the same gear tooth form, a very difficult method to produce and set up.

To set up you would have to have the drive roller with threads fully engaged in the threaded leadscrew, but at the same time the gear teeth of both screw and roller must also be engaged to ensure constant positional contact.

This would be a three roller set up, and as John pointed out you could have just one threaded roller, but cut with the gear teeth, and two other threaded rollers, no gear teeth, that are FLOATING end wise, with one of them eccentrically adjustable to make all rollers contact the screw closely.

It would work, but I don't think it's practical due to the amount of exacting extra work to make the threads engage as well as the gear teeth simultaneously, which with unhardened screws would not have a very long life span.

To eliminate backlash, the one roller that has the gear teeth, has thrust ball races at the ends.

This eliminates the problem of making each roller 1/3 positional spacing round the screw.

All the drive would occur on the one threaded and geared roller with ball thrust race ends.

I think it is much simpler to have the rollers just grooved, which is within the realms of the average DiY shop worker with a bit of lathe experience and a lathe, and you might get away with making the rollers out of brass or even alluminium, thus taking the wear factor away from the leadscrew as in conventional bronze nut practice, but whatever material is used you MUST have a ball thrust race at the ends of the drive roller, otherwise rapid wear will result on the end faces.

You could also make the two opposing rollers in the three roller set up spring loaded, so that the tension in the springs would ensure constant contact between drive roller and screw.
Ian.

hello,
I think it will be nice to see more video/pictures of realisation because this subject (conversion of rotation to linear motion) can be treated from the most simple to the most complex design.

It appear that nothing better than experimentation, it might take less time to do than to think sometime. It's absolutely the case here. So think less and build more, an approximative prototype will give you more informations than hours of neuronal computations, and you will see some problems desappearing or surge but in every case your brain will have to work on a much more narrow field and then be much more efficient .
Lucien

Dug around and found the little roller screw I made some time ago. Enjoy the video! I wish the video size limit were more like 2mb, but you'll get the picture.

it look like planetary roller screw this is a diy work ?
can you give some details of the making ?
it's the first planetary diy that i see!!!
very impressive

Sure. DIY, well "Did It Myself" ;-)
This was just a proof of concept toy.
I just made steel threaded bushings (1/4-32) and pressed them into the ID of 608ZZ bearings (skateboard bearings). I've got drawings somewhere of heavier, more practically applicable ones, but haven't found it yet.
There are *many* patents for this idea.
I think mine was simpler than most, but I haven't tried it on a heavily loaded case, which is where we'd probably find the need for more hardware!

Oh, no, not planetary. Female threaded rings on male threaded rod...

Nice work Mike,
I think I got it! Pure simplicity! If I understood it correctly, amazingly simple!

Well thought!

Best regards

Bruno

Nice work Mike,
I think I got it! Pure simplicity! If I understood it correctly, amazingly simple!

Well thought!

Best regards

Bruno
Thanks, Bruno. I think the more practical application would be straight grooves for 1:1 thread pitch use. I think someone here already said that.
Another interesting thing, and I'd say this is why simple bronze nuts are so good, is the lack of backdriveability.
Backdriving is all about efficiency. Roller screws and ballscrews can be easily back driven because of their high efficiency, so will not hold by themselves under cutting forces. Bronze in practical thread pitches is self locking.

Roller screws and ballscrews can be easily back driven because of their high efficiency, so will not hold by themselves under cutting forces. Bronze in practical thread pitches is self locking.

yes and this is very important for Z axis it keep your servo cold because you just use your servo for positioning and not to keep the position this mean also a better cut because no micro positionning
specially interesting for servo a bit too much undersized

Mike -could you clarify for my aged grey cells, what is the size/thread of the rod ?
In your video you say " 10 32", and I'm not sure what this is if it has to run inside 1/4-32.

Could you also clarify " the bearings, one in opposition to the other two". Is that thrustwise, or radially ?

Sorry to be so dim, but I'd like to understand what has excited Bruno so much. :)
John

Mike -could you clarify for my aged grey cells, what is the size/thread of the rod ?
In your video you say " 10 32", and I'm not sure what this is if it has to run inside 1/4-32.

Could you also clarify " the bearings, one in opposition to the other two". Is that thrustwise, or radially ?

Sorry to be so dim, but I'd like to understand what has excited Bruno so much. :)
John
Hi John, of course:
This little demo runs on a stick of 10-32 all-thread. That's #10 (.190" OD), 32 threads per inch. To get the differential action, it is important that the male and female threads be the same pitch, but different diameters. This particular combo, .190" running inside .250" gets you an approximate 3x multiplier on threads per inch. As these diameters approach the same diameter, the multiplier approaches infinity.
This demo has three identical plates with the bearing od sitting eccentric to the screw centerline. The center plate is rotated 180 degrees from the outer two, and light pressure applied radially inward, bucked by the two outer bearings, so middle is preloaded against the other two.
A more positive method of axial retention would be required for both the nut sleeves and the bearings if high loads or shocks will occur. This one is press fit.
I'm planning on drawing one up that's easily buildable. There's only two unique parts to it in my mind's eye, the roller nuts, and the bearing plates (two facing each other to capture each bearing.

What do we think a good lead screw spec would be? I'm thinking 3/4"-12(?) That's the screw on my Grizzly 10x22.

A quick thankyou Mike. Over on this side of the pond I'm not familiar with your #10 rod, but all is now clear.:cheers:

At least I was thinking correctly re the arrangement of the female threads, but I needed to be sure.
Regards
John

why does it not just spin on the bearings without moving linearly?what am i missing?

Mike,

Brilliant invention, when will the drawings be posted?

Very Kewl.

Jeff...

why does it not just spin on the bearings without moving linearly?what am i missing?
It is exactly like the male grooved rollers outside of the screw, which was the subject of this thread in which the inside of the roller bearings are held static and transmit the load to the carriage.
This way, the grooves are female, inside of the roller bearings, and the outside of these bearings are static and transmitting the load to carriage.
Clear as mud?
A drawing will speak a thousand words...

... I'm not familiar with your #10 rod, but all is now clear.:cheers:


Well 10-32 is English, after all... ha

i understand the subject of the thread...:) i am sure with a drawing i will get it...good work by the way.

Thanks! This is fun.
It occurs to me that finer than 8 threads per inch may not be useful, and as covered before, the differential multiplier will vary with load, which would be bad for stepper motor drive (it will not know where it is after some movement). The way for the differential to work is to have brush type motor, with linear encoder feedback. The benefit of this effective finer pitch is that no reducer is required at the motor end.

For stepper motor uses, we want a fixed pitch, so straight grooves are called for.
Assuming you want to run on a 3/4-8 screw, the easy way to make the straight grooves in the bearing liners (in absence of an NC lathe) will be to relieve a 3/4"-8 tap making it into a multi-groove tool and cut all grooves at once by moving the tool out radially. Nice that we don't have to be overly precision in the groove diameters, but being parallel to the centerline is a must, so all grooves share the axial load.
Obviously, the starting ID of the liners is larger than 3/4", like 1-1.25", and should be steel.

The contact patches at all the grooves is far superior, being concave against convex, vs. convex on convex for external, male rollers.

Thanks! This is fun.
It occurs to me that finer than 8 threads per inch may not be useful, and as covered before, the differential multiplier will vary with load, which would be bad for stepper motor drive (it will not know where it is after some movement). The way for the differential to work is to have brush type motor, with linear encoder feedback. The benefit of this effective finer pitch is that no reducer is required at the motor end.


Mike, can the variation be eliminated by adding pressure fro the nuts to the screw and/or offset the bearing by 120 degrees appart from each other instead of your actual setup? Maybe this could eliminate the slippage and insure a stable linear displacement?!?

Best regards

Bruno

To my way of thinking, the consistency of ratio (if that's what you mean by stability) isn't much to do about how the preload is accomplished. It just can't be perfect, and someone here pointed out earlier, if there's slip occurring on one roller, it's effective pitch is at 1:1, and the others are at something else. That's why the whole planetary synch jazz came into being, to keep it from jamming up.

Three at 120 degrees becomes 6 when you think about it. If you try to preload only three, no matter how you do it, since they are spaced along the axis, you end up twisting the nut assembly off axis. I did a six design, and there's a couple of ways to throw them in there so that (preload) forces balance. I'm currently in the KISS mode just using three at 180 degrees. Truly, you only have the strength of two, and the other is just there to balance preload force.

I'll post a solid design tonight so we have something, um, solid to talk at.

Well, this is about as simple as I can make it. The yellow plates have an eccentric cbore for the bearing. All six are identical, just flipped around appropriately. This one is for 3/4"-12 (I meant in my last post 12 tpi, not 8).
So to make one, you'd need:
(6) bearing plates
(3) #6905ZZ bearings, $5.95 from vxb.com
(3) ring rollers (one single point thread for the jam nut, straight grooves made with a butchered 3/4"-12 tap)
(3) jam nuts (custom turned and single pointed on this one)
(4) bolts n nuts
(1) tbd flange end plate with same bolt hole pattern

That was fun. I put off a ton of real work for this today! ;-)

that makes all the difference....i had something completely different...this makes sense.

Very clever design Mike!

With grooved rollers slippage causes no harm.
How is the tap butchered anyway?