Hi - I'm new to the board and new to CNC. I am converting a Rong Fu 45 table and column to a cnc machine. I have designed and built an anti-backlash screw consisting of two 1" threaded delrin blocks sandwiched between an aluminum piece that attaches to the table. One of the delrin blocks is bolted directly to the aluminum piece. The other delrin block rides on four 1/4" machine screws under spring pressure. In my experimentation using a digital caliper, I can't seem to get better than about 9-10 thous. backlash. And that is when the spring-loaded screws are really pretty darn tight. Is that normal? Will it be acceptable with software correction? I'm open to any suggestions experienced users might have.

I have attached a couple of photos showing my design.

Thanks!

Bob

I am not sure exactly how you set up to measure the backlash, so please forgive me for being "Captain Obvious". Perhaps the "backlash" is really in the screw supports. Do you have double-row, angular contact bearings installed on one end of the screw, or does it ride in radial bearings?

Another possibility is that when you apply force to the mechanism, the sprung-nut may be rotating a slight amount, manifesting itself as backlash. I doubt this is the case, since it is not a ballnut (capable of back-driving). Just a thought...

Good luck with this problem,

Rob

I would bet the sprung nut really is actually at least part of the problem. The sprung nuts bores appear to have raw thread passing through them, so it isn't going to be a tight fit. It isn't bushed or running on smooth shafting, so I'd bet there would almost have to be enough slop in there for it to be rotated slightly - thus getting rocked back and forth a bit by the leadscrew whenever it changes direction and drags the nut along for the ride.
Could be a number of other things as well, but that and bearings are a good place to start looking. 9-10 thou is a hell of a lot, you should easily be aiming at under 2.

I'm trying to figure out how your measuring the backlash with a caliper? Maybe on your design you could eliminate two screws and replace with alignment pins and let the springs ride on them.

Shouldn't the springs be on the inside of the block pressing out against the threads. I think you are working WITH the threading instead of against it.

The point to exerting pressure here simply ensures each of the two nuts remain firmly seated upon their opposing screw faces.
Either way should work equally well for this, but most nuts do indeed work through expansion and not compression.

I have a feeling this is more due simply to space and design requirements than any difference in performance, but if anyone knows differently I would be very interested to hear!

I'd go with spoiledbrat's view that what you are seeing here is end-float not backlash.

You cannot measure backlash accurately until you have counteracted any end-float in the system, or at least have measured it and can take it into account. For clarification 'backlash' is the amount you have to rotate the screw to get a change of direction of the table, converted back to linear form. For example, if you have a 10tpi screw and you can turn it 1/4 turn before the table reverses direction (as measured using a DTI) you have .025" backlash ASSUMING the screw doesn't move axially (along the screw). If you can push the screw and table assembly back and forth relative to the screw's mounting then that's end-float and needs to be counteracted by appropriate angular contact bearings or other approaches. End-float should be < 2thou before attempting to counteract backlash in the nut else you're on a hiding to nothing.

Thanks for all the info! First of all, you can see in the pic how I was measuring backlash using a digital caliper. With the assembly in place, I moved the screw in one direction, extended the caliper, and then turned in the opposite direction, watching both the caliper display and the graduated collar on the screw itself.

The screw has bearings only on the handle end and appear to be just two thrust type bearings, the inboard one bearing against a thick washer which, in turn, press against the screw itself.

Can someone explain what type of bearing would be ideal for this operation, etc.?

i agree that I should start first by making sure that any backlash problem stemming from the screw bearing should be eliminated.
Thanks!

Bob

Shouldn't the springs be on the inside of the block pressing out against the threads. I think you are working WITH the threading instead of against it.

I agree with this statement, the design pictured dose not take up any Backlash it just uses the Springforce against the second Nut.
Moving the Springs to the inside of the Blocks would improve the design, then you will be able to adjust the screws to take up the Backlash.
Still Nylon Blocks are not Ideal, they are a bit rubberish and when under load will increase friction and sticktion, so you may still end up with some Backlash.
But it would be far less then the 9-10 thou you are getting now.

I disagree with Torston. Your design does in fact work to eliminate backlash. I firmly believe that a spring need to be in the design when using non-precision screws, or any screw in an environment that a speck of dust can find its way to the screw. I use a similar setup, with an aluminum block tapped to 15/16-16, and a bronze block tapped the same. My ballnuts thread into them. There are SS guide pins that run through the bronze block for alignment. I set the "preload" (to me, that is the tension/compression between the nuts) to about 50 lbs. As long as my machining load is less that 50 lbs, I effectively have no backlash. Without the springs, an imperfection in the ball screw, or dirt in the ball screw could cause binding, or require a lot of torque to rotate the screw.

It truly does not matter wither you compress the nuts together, or push them apart. Hell, when the nuts start to wear, then use a mechanism to push them apart. Just like new!

Rob

Spoiledbrat and stepper monkey are correct. I just keep thinking in the way a screw and nut work that comes naturally to me - turn the nut on the screw then the nut moves and the screw remains in one place. The other way it's just the opposite - you turn the screw then the nut stays and the screw moves up and down.

:cheers:

Rick

All that matters with an a/b nut is that one nut is firmly bearing on one screw face and the opposing nut is firmly bearing on the opposite face. The relative position of the two is irrelevant as long as they are on opposing faces.
I think the layout is fine either way, compression or tension, you just need to keep the nuts from moving axially relative to one another (i.e. bushes or guide pins), and to remove any end float. I would suspect there is a little of both going on here, but the basic arrangement itself is sound.

I expect your probably is probably in the duel contact bearing and the end mount. Take a dial caliper with a magnetic mount and put it on th eend of the screw ( where the crank handle is) . Turn the screw one way until the backlash is out and the set your diel indicator to zero, then turn it back the other way and see if the screw moves the dial indicator. If it does the slack is in the bearing mount and you may need to shim it until you get it all out.


I wish I had a camera because I have the same problem with my little X1 and it will take making some bushings and shims to fix it. If I tighten the crank handle nut all the way up it will lock the screw up so it needs something that will stop the handle at one point when you tighten the nut up. Once this is done then shiming to take the slack out can be done.

Take your crank handle and bearing support apart and look at the design very close and you will see what I mean. Your preloaded screw nut setup will work as long as the sprung nut doesnt try to turn . If so it will need some pin guides so it cant try to turn. The springs alsdo have ot be strong enough to not give when cranking the screw but you should not be having thta problem just trying to move the table.



Jess

The current design wil wear the nuts unecessarily as there will always be a load on the threads imposed by the springs. The design will only compensate for lash until the cutting force exceeds the spring force opposing the moveable nut, and then all of the slop (backlash) in the fixed portion of the nut will be introduced into the tool path. The moveable nut needs to be help precisely in place by the locating screws.
Joe

Bob,
I am thinking about making some antibacklash blocks for my mill too. I was also considering delrin as a material to replace the cast iron ones that came with my mill. The cast iron wears so fast- though it is a new mill and I have only been using it manually, and perhaps the blocks are still wearing in.



About your design- I agree with Hackmax- the springs should be on the other side. As it is, when the table is under heavy load the springs could compress.
Looks good though. Did you cut the thread in a lathe?


Does anyone have a material they would recommend over delrin for antibacklash nuts?

Stefan

Thanks for all the info! First of all, you can see in the pic how I was measuring backlash using a digital caliper. With the assembly in place, I moved the screw in one direction, extended the caliper, and then turned in the opposite direction, watching both the caliper display and the graduated collar on the screw itself.

The screw has bearings only on the handle end and appear to be just two thrust type bearings, the inboard one bearing against a thick washer which, in turn, press against the screw itself.

Can someone explain what type of bearing would be ideal for this operation, etc.?

i agree that I should start first by making sure that any backlash problem stemming from the screw bearing should be eliminated.
Thanks!

Bob

Hi Bob. What you have in the pictures is a "torrington" bearing on either side of the casting, plus a bearing or bushing inside the casting to take the radial forces of the screw. That is a very good setup. Most people use an angular contact bearing instead becuase it takes both the axial and radial loads in one bearing and is cheaper and easier to manufacture. It is also normal to have bearings on one end of the screw only, especially on a short screw.

The thrust bearing "torrington bearing" normally has a hardend washer on both sides of the balls to prevent the non hardened parts from wearing.
This hardend washer is actually the bearing race.

I would start by adjusting the preload on the torrington bearings to about 10oz/in above what it takes to rotate the handle before preload. I have a small bar type torque wrench for this. This should guarante that there is no axial movement of the screw itself.You can verify this by locking the table in place and then measureing if the screw moves in and out of the casting/bearings while turning the handle with the kind of force seen during machining. DON"T OVERDUE IT. You will have to setup a dial indicator on the end of the handle or find someother way to measure this as we haven't eliminated the backlash of the screw itself Yet and you can't use the graduated collar for this measurement.

The current design will wear the nuts unecessarily as there will always be a load on the threads imposed by the springs. The design will only compensate for lash until the cutting force exceeds the spring force opposing the moveable nut, and then all of the slop (backlash) in the fixed portion of the nut will be introduced into the tool path. The moveable nut needs to be help precisely in place by the locating screws.
Joe

Extra wear of the nuts is a normal byproduct of using a spring loaded nut system. It is also normal for this setup that if the cutting forces exceed the spring loading it will introduce more slop into the tool path. Springs should be stiffer than the max cutter force predicted.

The other option is to hard mount the 2 parts of the nut and shim until backlash is minimal. However, you will never get "0" backlash as the screw is not perfect. You would have to set it to zero on the tight parts of the screw and live with whatever the backlash is on the loose parts of the screw.

Bob,
I am thinking about making some antibacklash blocks for my mill too. I was also considering delrin as a material to replace the cast iron ones that came with my mill. The cast iron wears so fast- though it is a new mill and I have only been using it manually, and perhaps the blocks are still wearing in.

About your design- I agree with Hackmax- the springs should be on the other side. As it is, when the table is under heavy load the springs could compress.
Looks good though. Did you cut the thread in a lathe?

Does anyone have a material they would recommend over delrin for antibacklash nuts?

Stefan

The way this works is to have one part of the nut take the load in one direction and the other part of the nut take the load in the other direction. Doesn't make a difference if the parts are in tension or compression. In a elephant sized machine it might make a difference but in this size of a machine it makes no difference. For higher forces than the Delrin is happy with I have heard that bronze is a great canidate. If you are spring loading the nut, lubrication is very important, not to mention cleanliness. Excelerated wear, over what is considered normal for the same nuts without the springs, is part and parcel of theis design. I suggest that as you are making the nuts, make several sets, so when you need to replace them, they are on hand.

I would start by adjusting the preload on the torrington bearings to about 10oz/in above what it takes to rotate the handle before preload. I have a small bar type torque wrench for this. This should guarante that there is no axial movement of the screw itself.You can verify this by locking the table in place and then measureing if the screw moves in and out of the casting/bearings while turning the handle with the kind of force seen during machining. DON"T OVERDUE IT. You will have to setup a dial indicator on the end of the handle or find someother way to measure this as we haven't eliminated the backlash of the screw itself Yet and you can't use the graduated collar for this measurement.

TOTALLYRC - thanks for your advice and description of the bearings. I don't see a way to adjust the preload on these bearings - there is no take-up type of nut on the end of the screw itself - am I missing something here?

Thanks!

Oh, BTW to S. Sharpe - I did not cut the threads on the lathe. I made two taps (RH & LH) based on the thread profile of the screw and its major and minor diameters. I then held the tap in the lathe chuck and rotated it at the lowest speed setting while keeping the delrin block from rotating using a pipe wrench. It actually worked out quite nicely yeilding a very good fit.

Bob

I would start by adjusting the preload on the torrington bearings to about 10oz/in above what it takes to rotate the handle before preload. I have a small bar type torque wrench for this. This should guarante that there is no axial movement of the screw itself.You can verify this by locking the table in place and then measureing if the screw moves in and out of the casting/bearings while turning the handle with the kind of force seen during machining. DON"T OVERDUE IT. You will have to setup a dial indicator on the end of the handle or find someother way to measure this as we haven't eliminated the backlash of the screw itself Yet and you can't use the graduated collar for this measurement.

TOTALLYRC - thanks for your advice and description of the bearings. I don't see a way to adjust the preload on these bearings - there is no take-up type of nut on the end of the screw itself - am I missing something here?

Thanks!

Oh, BTW to S. Sharpe - I did not cut the threads on the lathe. I made two taps (RH & LH) based on the thread profile of the screw and its major and minor diameters. I then held the tap in the lathe chuck and rotated it at the lowest speed setting while keeping the delrin block from rotating using a pipe wrench. It actually worked out quite nicely yeilding a very good fit.

Bob

Take a look at the picture in post #8. There isn't any threaded portion to make the adjustment. It is possible it is in the parts that are not in the picture. I would take a look at the owners manual and see if they offer any insight into how this is adjusted. It might be part of the handle assemble or the graduated collar assembly. It could even be spring loaded by a wave washer or some other non threaded means. Let me know if you can't find anything.
Mike

Mike - I've looked at all the parts as well as an exploded view diagram of the table and there does not appear to be any type of adjustment for the table screw to take out any end play - suggestions?

Bob

Mike - I've looked at all the parts as well as an exploded view diagram of the table and there does not appear to be any type of adjustment for the table screw to take out any end play - suggestions?

Bob

I have attached your picture with an arrow drawn on it. Take a picture of everthing that goes where the arrow is. It should be the graduated dial, handle and stuff. The order of parts, I am guessing, should be another washer/bearing race, graduated dial, possibly some spacers and the handle.


If you are going to not use the dial, handle, and such on this end, the easy way to set the backlash could be as simple as a shaft collar. Put the bearing race on, then the collar. Push the collar up against the race as hard as you can without getting silly, and have a friend tighten the set screws. The right way would be to thread the end of the screw and use a double nut system, one to set the backlash and the other to lock it in place.

Mike

For higher forces than the Delrin is happy with I have heard that bronze is a great canidate. If you are spring loading the nut, lubrication is very important, not to mention cleanliness. Excelerated wear, over what is considered normal for the same nuts without the springs, is part and parcel of theis design. I suggest that as you are making the nuts, make several sets, so when you need to replace them, they are on hand.

Hi, TOTALLYRC and everyone

Do you have any comments or experience on wearing resistance of Delrin and bronze nut?

Drift

Hi, TOTALLYRC and everyone

Do you have any comments or experience on wearing resistance of Delrin and bronze nut?

Drift

No actual personal experience. I am lucky enough that my router has ball screws on the y and z and rack/pinion on the x. My bport is all ball screw and I havn't run my Taig long enough to wear anything out.

Mike - as you can see in the picture, the only pieces that go on the screw end are the hardened washer for the bearing and the "clutch" type casting which engages the handle. The graduated dial simply goes over the inboard end of the "clutch" casting.

I was thinking of designing a threaded sleeve that would be pinned to the shaft and would allow preload adjustment by a nut around the threaded sleeve.

Thoughts?

Bob

Mike - as you can see in the picture, the only pieces that go on the screw end are the hardened washer for the bearing and the "clutch" type casting which engages the handle. The graduated dial simply goes over the inboard end of the "clutch" casting.

I was thinking of designing a threaded sleeve that would be pinned to the shaft and would allow preload adjustment by a nut around the threaded sleeve.

Thoughts?

Bob It looks like the clutch casting is how the backlash is adjusted. It is simply pushed up against the washer/bearing race and tightened
It is a very basic way to do it. I like the idea of a threaded adjuster to preload the bearing. Pinning it to the shaft also sounds good to me.

I haven't been using Delrin nuts long enough to kill any, but from what I've seen from the router guys postings at least, they think you apparently can't kill them. I've seen mention that they can last longer than bronze even, but I don't know the specific application they are putting them to so I can't say with any certainty.
I have made some white Delrin nuts as "temporary" units for a prototype machine, but they have held well enough so far I am not looking to replace them.
I have also recently gotten some strange "drops", or off-cuts, from a research lab that was using a lubricant impregnated version for thier slides. The stuff is not white - it's brown, slick as all get out, and I don't know anything beyond that, as I haven't found an excuse to use it yet. Anyone know what this stuff is called exactly or how well it might work for a/b nuts?

hello,
I look at the rong fu

/http://www.fignoggle.com/machines/rf45-dovetail-square-column-mill/enco-rong-fu-45-dovetail-column-mill-first-look.htm

I do not think that delrin is a good idea .Delrin is perfect for wood machine but this is for hard metal and if you work by hand backslash is not a concern elasticity YES
if you work with cnc baclslash,backdrive,elasticity and temperature of the nuts and then of the screw are problems that should be adress.
ACME nuts:
avantage :it's small section,it's cheap resistant and if of good quality precise to very precise also no backdrive the backslash can be address by using 2 nuts separate by a spring eg belleville (this will increase the amount of energy taken by the nuts

desavantage only one:the amount of energy taken by the nuts and restitute under heath (up to 70% so fast wear elongation of the screw bigger servo etc..) mean that under high load very small speed can be used (with lubricant) and that's why it is used for manual tools BUT it work fine.

ballscrew :
avantage :it's a very small section with 2 nuts it's a no backslash device ,the amount of energy taken by the nuts is minimum (<5%)and this under high load.a single nuts ballscrew can be nobackslash using bigger balls this will increase the amount of energy and thus the wear of the screw.Easy and cheap to make with a very expensive equipment.
disavantage: it's fragile a crash will square your ball if not destroying the bed of the screw,it's expensive to VERY EXPENSIVE depend of the size in diameter ,there is a lot of backdrive (this make a machine equiped with ballscrew non operable manually !!!) and unless you have the very expensive equipment you cannot do it yourself .A real no backslash (double nuts) willbe more expensive.

THERE IS A DIY ALTERNATIVE......the annular rollernuts

Avantage:no backslash,no backdrive,resistant can be Do It Yourself, do not need to change the original screw amount of energy taken by the nuts small <>20% cheap to affordable (depend if it's for 20kg or 500kg machine) it's a very high quality level solution (better than ball screw)
Disavantage:big section (so real problem of implementation on existing machine) cannot be bought on a shop and if it was it will be much more expensive than ballscrew due to the amount of work
here some realisation
and the (very long)thread in cnczone
http://www.cnczone.com/forums/showthread.php?t=13593&page=51
Lucien

While the use of a digital caliper is clever I can't help but suspect that you won't get accurate, repeatable results with it. The way it is positioned it will be able to move just slightly and since it is not spring loaded you will no doubt wiggle it a bit when trying to stretch it back out. Also, it it resting against a rough casting on one side so it would not take much movement to significantly alter your reading.

You really need a dial indicator and a good (magnetic) base for it. I like to use a cheap(er) 0-1" 0.001" graduations (dual dial) until I'm sure everything is in the ballpark. (This type is also good for calibrating you step/unit later). Then I'll move on to a nicer 0-15-0 (0.005 graduations) or 0-4-0 type (0.0001 graduations) to measure things down to a gnat's a$$.

The tip of your dial indicator needs to rest on a machined surface, not a rough casting. You also need to be sure that your indicator is positioned so as to be square to the travel your measuring. Good indicators are expensive but Enco has some decent Enco branded ones for less than $70. While they are not a super top quality unit they will serve you MUCH better for this task then a caliper will.

Hope that helps...

This is how I did mine on both my router and my new lathe.
This is a much larger spring. Those smaller springs that you are using can't provide the same amount of pressure that a larger one would. With the weight of your table and cutting forces as well as G force when changing direction, you really want the nuts to stay apart from each other. If they don't, then you have just introduced some backlash.
In this design, the lead nut is thread into a steel angle that is the mount. I lock it in with a set screw.
I mill a U channel shape to fit around both nuts. I lock this channel to the lead nut. Then load up the second onto the spring. Then the channel holds it and keeps it from turning, but it is still able to slip back and forth to remove any backlash.
I am not able to measure any backlash with this design. It has to be smaller than .001".
That is pretty good for a ball screw with a rating of .004" average per foot.

http://www.cnczone.com/forums/attachment.php?attachmentid=72462&d=1230561899

Lee,

Have you by any chance documented your build somewhere?

I'd love to see more if you wouldn't mind? :)

Thanks,
Dave

This is a much larger spring. Those smaller springs that you are using can't provide the same amount of pressure that a larger one would. With the weight of your table and cutting forces as well as G force when changing direction, you really want the nuts to stay apart from each other. If they don't, then you have just introduced some backlash.

well belleville is a cheap and perfect solution in this case it will reduce the length of the assembly you can have the amount of force you want by adding belleville
here a useful link for this mid-nineteenth century patent
http://www.bellevillesprings.com/3-18.pdf

Here is a link to all three of my builds.

Router. (http://www.cnczone.com/forums/showthread.php?t=8632&page=1&pp=40)

Mill. (http://www.cnczone.com/forums/showthread.php?t=32609)

Lathe. (http://www.cnczone.com/forums/showthread.php?t=59421)


I did try belvilles and wavy washers initially. I found they were much more difficult to adjust correctly. This large spring of course, does take up quite a bit more space, but you can apply exactly as much pressure as you want without removing the ball nut so many times to try and get the belvilles just right.

I did try belvilles and wavy washers initially. I found they were much more difficult to adjust correctly. This large spring of course, does take up quite a bit more space, but you can apply exactly as much pressure as you want without removing the ball nut so many times to try and get the belvilles just right.


since you know the ideal force you need (by measuring the length of your spring) it's easy to choose the corect belleville assembly.the internal diam. is determined by the leading screw and the thickness by the force needded and the arrangement you choose serial or parallele

Here is a link to all three of my builds.

Router. (http://www.cnczone.com/forums/showthread.php?t=8632&page=1&pp=40)

Mill. (http://www.cnczone.com/forums/showthread.php?t=32609)

Lathe. (http://www.cnczone.com/forums/showthread.php?t=59421)


Thank you very much, sir-

Dave

Hey Lee,
I watch your realisation .
since definitely you are CNC ADDICT (I now the feeling) make some economy next time for your next realisation do not use ballscrew anymore you are very well equipped to make your own roller screw

Here is a link to all three of my builds.

Router. (http://www.cnczone.com/forums/showthread.php?t=8632&page=1&pp=40)

Mill. (http://www.cnczone.com/forums/showthread.php?t=32609)

Lathe. (http://www.cnczone.com/forums/showthread.php?t=59421)


I did try belvilles and wavy washers initially. I found they were much more difficult to adjust correctly. This large spring of course, does take up quite a bit more space, but you can apply exactly as much pressure as you want without removing the ball nut so many times to try and get the belvilles just right.

Commercial double-nut ballscrews often use Bellevilles - they're easy to adjust either by design or by shimming. You should know exactly what pre-load is appropriate (about 180-200 pounds for a 5/8" ballscrew, IIRC) as that is specified by the nut manufacturer, and your application. I used wavy washers on my knee mill conversion, and they work great.

Regards,
Ray L.

The problem I was having with the belvilles was when the right amount of tension was applied, the nut was not close to vertical like the other. It took many tries and I still could not get them to line up right in order to use the channel I had made. It is a crucial element. It keep the second nut from turning any, but does allow the spring to do it's thing.
Belvilles when done right will fit a much tighter space for sure. These big springs are just so much simpler. I was not tight on space either, so it worked fine for both the router and the lathe. My mill has much better ground ball screws and didn't even need anything like this. They are preloaded with a big ball and then a little one and so on. They are very close to perfect.

The problem I was having with the belvilles was when the right amount of tension was applied, the nut was not close to vertical like the other. It took many tries and I still could not get them to line up right in order to use the channel I had made. It is a crucial element. It keep the second nut from turning any, but does allow the spring to do it's thing.
Belvilles when done right will fit a much tighter space for sure. These big springs are just so much simpler. I was not tight on space either, so it worked fine for both the router and the lathe. My mill has much better ground ball screws and didn't even need anything like this. They are preloaded with a big ball and then a little one and so on. They are very close to perfect.

That's where the shims come in. If you need the nuts in a specific orientation, you use shims provide the room needed to bring them around. On mine, I simply designed the attachment so the correct spacing was present between the nuts when they were properly aligned. The compressed thickness of the bellevilles, or wavy washers, required to create the desired pre-load can be easily determined from their specs.

Regards,
Ray L.

hello,
A little remind about belleville
You should consider belleville as a complex part who need calculation that's the meaning of the link i give previously.
The strategy of dichotomy (one long shot, one short shot and hopefully one in) should be use at last ressort since often the belleville is implemented in complex assembly .shim are important part for the use of this old but very efficient device

Rokag3 - I have made the delrin nuts and have gotten down to about 0.0015 backlash, but this is under no load and with considerable force to turn the lead screw.

I really like your design with the annular nuts. In your drawing, you show an eccentric adjustment - is this to adjust the parallel distance between the bronze annual nut and the leadscrew? Do you also bias the annual nuts in opposite directions aliong the direction of travel?

Thanks!

Bob

In your drawing, you show an eccentric adjustment - is this to adjust the parallel distance between the bronze annual nut and the leadscrew? Do you also bias the annual nuts in opposite directions aliong the direction of travel?

|
the roller must roll ! if it does not roll then the roller screw will act like a regular acme screw .in order to roll each roller should be in contact with the lead screw it's also very important to assume the no backslash feature since at least one ring of the roller should be in contact with the opposite edge of the screw .ideally each ring should be tangent with both edge of the screw in practice it's probably not the case but the number of rings make it true statically. for this reason the role of the excentric (on each roll and on both side) is major it keep a perfect contact with the lead screw. also since there is a maximum force that can be applied to the roller to avoid blockage of the roller we are using a spring to keep an even pressure and make the system self adjustable.
The excentric can also be use ase a safety feature to free the load from the screw in case of crash.
It's not difficult to do . If you have your mill working on cnc it's very easy .Use a lot of lubrifiant and a bigger demultiplication with your delrin nuts it will be slower but it will work

but the number of rings make it true statically
read statistically from statistic and nothing to see with static.
about the bias question the profile of the rings is slightly assymetrical (it's the projection on a plane of the screw profile with an angle equal to the helix angle)what is important is that you must open the angle of the profile . If you have a male angle of 45deg and you present it to a female form of a corespondant deep both edge of the male will be tangent to the female so it's a lateral pressure that make the job and not an axial

Bob,

At the risk of sounding like an evangelist;
As Lucien has pointed out my Roller Nut design (http://www.cnczone.com/forums/showthread.php?t=13593&page=37) should fit straight on your RF40 (I have an RF30 clone).

If you have a lathe, about £30 or $30 for bearings and some scrap brass & aluminium, you should be able to knock them up in a week or so.

I can let you have drawings if you're interested (I promised I'd post them a while ago).