[bigspike]

I know that this has been widely discussed, but I was just about to order ball screw stock from Roton and did a new search on the forums to recheck my decision only to question the lead error and quality level of the Roton ball screws.

Some members swear by them as well as the eBay, Chinese C7, homeshopcnc offerings. Others would prefer ground ball screws in a C5 minimum quality. Roton was quoted as having .009"/foot max lead error, another member replied with a spec of .003"-.006" for the Roton. Nook specs a .004/ft for their SRT line.

As I understand it the error is not necessarily progressive but may in fact vary from one inch to the next.

Mach3's & EMC2's screw mapping functions are mentioned but very little information is available about using them. Such as:
What DRO will output the data to a PC for direct interpolation?
Or is this even possible?
And if it is can anyone reasonably afford it?

One post I found described using a digital scale with a built in display and from what I understood the procedure is as follows.

1. secure the scale to the axis
2. secure the slider to the machine
3. home the axis
4. jog in Mach3 one inch (the post questioned should a smaller or larger increment be used - no reply was ever left)
5. enter the actual movement in the screw mapping window in Mach3
6. save the correction

then what?
repeat from home for succeeding inches, i.e. 2", 3" etc?
If the scale cannot be interfaced with the PC, how does one generate the screw mapping correction curve to be displayed in the window?

I don't need machine resolution to .000001" but don't think I can live with .006-.009"/foot without repeatable error correction.

If I can get more information regarding the potential for screw mapping at a reasonable cost. Hopefully using an inexpensive digital scale or DRO that I can then install on my budget lathe. Then I can proceed with the Roton screws and finally move forward.

Additionally, if anyone has a current cost & quality analysis of 5/8"-3/4" ball screws and nuts for Roton vs other brands it may be more cost effective to purchase better quality ball screws with a lower rated lead error than to attempt screw mapping with it's associated scale/DRO cost. I am very hesitant to use the Chinese ball screws due to the more than likely possibility of incorrect specifications, poor quality.

I am not having much luck getting quotes from Nook, Thomson, Bosch-Rexroth, Rockford, etc for competitive analysis. I also prefer to buy from a US distributor rather than pay for international shipping and the longer wait time. If I ever get this info, I wil post it here.

Sorry for the long convoluted post, but there is no definitive source for all this information. For those who feel they need to know what I plan to mill...assume everything and anything. I never know where my interests will go.

FYI this is for an RF45 type machine and I plan to machine the ends myself

[691175002]

There are many tradeoffs to consider since different solutions have different advantages and disadvantages (cost, backlash, accuracy, thermal error etc...).

I decided that 5 micron glass scales (~160$ per axis) and c7 screws were an efficient combination, however there is very little information available.


I should have c7 screws and glass scales operating in a week or so which might make things more clear.

[bigspike]

There are many tradeoffs to consider since different solutions have different advantages and disadvantages (cost, backlash, accuracy, thermal error etc...).

I decided that 5 micron glass scales (~160$ per axis) and c7 screws were an efficient combination, however there is very little information available.


I should have c7 screws and glass scales operating in a week or so which might make things more clear.

Sure, but how do you connect the scales to Mach3? What brand, supplier?

And do you have any more info on mapping? Or will you report back when you get them setup?

[SCzEngrgGroup]

Accuracy comes not from choosing one correct component, but from a thorough analysis and design of the entire system. The best, most expensive, ballscrews in the world will not turn a cheap Chines mill into a high-precision machine, as there are many other sources of error that are at least as significant as the ballscrew accuracy. Anyone who claims to have achieved resolution and accuracy (and those are two VERY different things!) of anything less than several thousandths on one of these machines with an open-loop control system (i.e. - steppers or low-cost servos) does not understand how resolution and accuracy are really spec'd and measured in the real world. Just SOME of the other MAJOR factors that MUST be considered are:

1) Basic machine geometry - cheap Chinese mills are typically quite poor in this respect, with errors of 0.004"/foot not at all uncommon. Combine a few linear errors, and a few perpendicularity errors, and you have a real mess.
2) Machine rigidity - cheap Chinese mills are typically quite poor in this respect. Deflections of several thou under relatively light loads are not at all uncommon. There's a reason high-precision machines weigh many tons.
3) Spindle runout, and it's close cousin tool/toolholder runout. Plan on having your spindle re-ground, and replacing the spindle bearing with expensive, high-precision bearings to deal with this one. In many cases, re-engineering of the whole spindle assembly is required.
4) Axis drive systems - Anything less than a double-close-loop servo with both position (via linear encoders on all axes) and velocity (via rotary encoders on all motors) is fundamentally incapable of achieving anything much better than a few thou accuracy. If, on a given day, the machine DOES deliver such accuracy, it due as much to luck as anything else.
5) Perhaps most significant - differential thermal expansion of the machine and it's components. This can be one of the biggest contributors to overall system error, and even operating in a temperature-controlled environment will only partially protect you. Without #4, all the screw-mapping in the world will do nothing to protect you from this MAJOR source of error. And, BTW, the screw-mapping as currently implemented in Mach3 is completely useless, which is why nobody actually uses it.
6) Stiction - Do you plan to retrofit all axes with Turcite or high-quality linear, ball-bearing guides? If not, stiction can easily add a thou or two of uncertainty to any particular move.

Regards,
Ray L.

[bigspike]

We all know that this is not a Haas mill...
And one issue at a time.

If in a worst case installation, I use the Roton ball screws and get .009"/ft lead error. Then I could have an error of .018" (nearly 1/32") on a 24" part which is completely unacceptable for metalworking. It may even be to much for wood.

If I use high quality ground ball screws at .001"/ft then the error is only .002" in 24".

I understand the limitations of this type of machine and many of the issues you relate will be addressed. I just want to find a happy medium between accuracy & budget. With perhaps a method of compensating for lead accuracy or better quality ball screws.

The mapping in Mach3 may be useless, but I can find no information from anyone who has any understanding of how it was supposed to work. If no one knows how to use it, then it would be useless. There is nothing in the Mach3 manual either.

EDIT: Of course if the error was cumulative in a even amount/inch then a steps per inch adjustment would suffice. But if it varies along the screw, then that would not work. If the .009" error was to fall within the 6" to 11" range of a ball screw... and again in the 15-20" section.

[bigspike]

I just found this

Some of the cheaper screws on the market don't have any calibration or class details when supplied. They seem to be consistent over length, the same pitch along the length of the screw, but often a little out to the pitch you expect them to be, probably why they are cheaper.
That's not a problem with most CNC packages you can easily measure a long length and adjust the steps/mm to account for the error. I often use a long digital vernier, and get the spindle to push the jaw along zero it then push it some more so avoiding any backlash

Dr. Mark Vaughan Ph'D., B.Eng. M0VAU

*****************************
If adjusting the steps per inch will make it right, that is the way to go. If the screw is inconsistent, you can use screw mapping to adjust out the error.

Ron Thompson

*****************************
Screw mapping gets a lot more complex though.

There are three main types of error, a continuous error, easily taken care
of with steps/inch and common on cheaper screws.

Then sometimes you get an error along the screw length where the pitch
changes with respect to length. This can be taken care of with screw mapping, though the measurements get a bit tedious without a something like glass slide to pull data from.

Sometimes the error changes around the screw so you get a change with respect to rotational angle, This one isn't so easy to allow for, it creates a sine wave modulation upon the main screw mapping chart, but the screw angle then needs to be known when the machine homes since that becomes critical. This is usually that small you don't worry about it, and if you do need to you buy better screws, unless it's an instrumentation need where better screws may not exist.

Thankfully screw mapping is rarely required, I'm not sure I even know anyone in mach using it, or even if it actually works. I've played with spindle speed mapping and gave up as that didn't work at that time.

Dr. Mark Vaughan Ph'D., B.Eng. M0VAU

[691175002]

I do not believe mach3 is properly equiped to make use of leadscrew mapping or linear scales. I will be using kflop.

If I were to try measuring pitch error, at a minimum I would use a 4" gauge block and a tenth indicator. The chances of a pair of calipers measuring more accurately than a ballscrew is unlikely. Most people can't even get their calipers to repeat within a thou.

Ray is correct in saying that even the worst screws will be relatively small compared to other sources of error.

[Mad Welder]

If cost is your main concern you won't go far wrong with either the C7 or Roton ball screws regarding your question of an analysis chart Hoss actually has a very good chart put together he refers to it a couple of times in his thread and it's on his website free download also sorry I can't get my hand on the specific link address but if you post that question in Hoss's thread he'll pull that link out of his hat in a second and again Ray's bang on accurate in his analyses and I learned that too after machining parts and not getting the precise measurements that I had drawn and scratching my head for ages because my screw backlash was measuring minimal and to add to my confusion I was getting near excellent square pockets but was losing tolerances while machining curved surfaces when all 3 axis were moving in unison. And it wasn't until reading a similar answer in another thread where Ray had posted pretty much the same as here where I found my answer........

Hopefully this is of some use......

[wizard]

Accuracy comes not from choosing one correct component, but from a thorough analysis and design of the entire system. The best, most expensive, ballscrews in the world will not turn a cheap Chines mill into a high-precision machine, as there are many other sources of error that are at least as significant as the ballscrew accuracy.

This is certainly true.

Anyone who claims to have achieved resolution and accuracy (and those are two VERY different things!) of anything less than several thousandths on one of these machines with an open-loop control system (i.e. - steppers or low-cost servos) does not understand how resolution and accuracy are really spec'd and measured in the real world.

This MAY not be the case. Some guys put a lot of effort into their conversions and I don't doubt that a few actually do get the values quoted. Having been involved in some fairly significant rebuilds of old American hardware I can state without embarrassment that worn out machines can easily be far worst than some of the hobby machine coming new from China.

Just SOME of the other MAJOR factors that MUST be considered are:

1) Basic machine geometry - cheap Chinese mills are typically quite poor in this respect, with errors of 0.004"/foot not at all uncommon. Combine a few linear errors, and a few perpendicularity errors, and you have a real mess.

Yes, this can be seen in the threads here and in various other forums. However not every machine is the same mess and some times one does get lucky. More importantly one can improve these machines significantly with a little effort. Being small makes for an easier rebuild.

2) Machine rigidity - cheap Chinese mills are typically quite poor in this respect. Deflections of several thou under relatively light loads are not at all uncommon. There's a reason high-precision machines weigh many tons.

No one I know of is surprised by this. Further judicious use of CNC control,can minimize that working deflection. The reality is that if used as intended, frame stiffness can be worked around. The key is to avoid thinking like a production machinist with a hot date sitting in the parking lot.

3) Spindle runout, and it's close cousin tool/toolholder runout. Plan on having your spindle re-ground, and replacing the spindle bearing with expensive, high-precision bearings to deal with this one. In many cases, re-engineering of the whole spindle assembly is required.

Even here it is not a given that a specific spindle will be bad. Further buy a model or two upscale and you do get better hardware. This stood out sharply when I went drill press shopping at Grizzly. The lower end models had terribly loose spindles and extremely sloppy quils. This lead me to going up scale to a larger machine, yes it cost more but is far better constructed.

I think the key here is to know what you are buying.

4) Axis drive systems - Anything less than a double-close-loop servo with both position (via linear encoders on all axes) and velocity (via rotary encoders on all motors) is fundamentally incapable of achieving anything much better than a few thou accuracy. If, on a given day, the machine DOES deliver such accuracy, it due as much to luck as anything else.

This is pure garbage. I've worked on diamond turning hardware generating optical surfaces that didn't have all the modern feature you seem to think are needed. Velocity feedback was via a old fashion tach and the position feed back via an extremely high precision rotary encoder. At the time castiron ways ruled the day. The machines could generate radiuses withing a couple of Microns all day long.

5) Perhaps most significant - differential thermal expansion of the machine and it's components. This can be one of the biggest contributors to overall system error, and even operating in a temperature-controlled environment will only partially protect you. Without #4, all the screw-mapping in the world will do nothing to protect you from this MAJOR source of error. And, BTW, the screw-mapping as currently implemented in Mach3 is completely useless, which is why nobody actually uses it.

Haveing rerouted a few air conditioning ducts in my time I can't dismiss the relaity of thermal expansion. However a machine will reach an equilibrium at some point.

6) Stiction - Do you plan to retrofit all axes with Turcite or high-quality linear, ball-bearing guides? If not, stiction can easily add a thou or two of uncertainty to any particular move.

An agressive lubrication system helps here. Further that lubrication system should be combined with some sort of axis slewing sub routine to maintain the effectiveness of the lubrication system.

Regards,
Ray L.

For the most part I don't disagree with you overall. You do have an excessively narrow interpretation here. Yes the tolerance on these mills is wanting, but it is not something that a skilled person can't over come.

Really it would be great if wee could all afford a Hass or similar industrial mill in our cellars. Most of us can't so we look for alternatives.

[SCzEngrgGroup]

This MAY not be the case. Some guys put a lot of effort into their conversions and I don't doubt that a few actually do get the values quoted. Having been involved in some fairly significant rebuilds of old American hardware I can state without embarrassment that worn out machines can easily be far worst than some of the hobby machine coming new from China.

Taking a single measurement against a 1-2-3 block and getting an impressive number under ideal conditions does not in any way indicate the machine is REALLY accurate in operation. AFAIK, the ONLY widely accepted, meaningful test of machine accuracy is the "ball test". I've NEVER seen anyone here run that test on their machine - no doubt because the equipment to do it costs thousands of $. And, until someone does run this test, any accuracy claims they make are just wishful thinking and cherry picking of best-case data which does not in any way reflect the true machine performance when making real parts.

This is pure garbage. I've worked on diamond turning hardware generating optical surfaces that didn't have all the modern feature you seem to think are needed. Velocity feedback was via a old fashion tach and the position feed back via an extremely high precision rotary encoder. At the time castiron ways ruled the day. The machines could generate radiuses withing a couple of Microns all day long.

Pure garbage? I'd like to see the references on which you base such a blanket statement. Without credible references, you're simply giving your opinion, which may be based on nothing more than wishful thinking. Is this your area of expertise? I can cite several good references from credible sources to backup my statements, this being one of the more easily digested ones:

http://www.heidenhain.de/de_EN/php/d...-0209/file.pdf

In particular:

"In order to prevent heating of the ball screw and the surrounding parts of the machine, some ball screws feature hollow cores for coolant circulation. In semiclosed-loop operation the positioning accuracy is affected by thermal expansion of the ball screw and thus depends on the temperature of the
coolant. A temperature increase of only 1K results in positioning errors up to 10 μm over a traverse range of 1 m. Common cooling systems, however, are often unable to restrict the temperature variations to values significantly below 1K.

For drives in semiclosed-loop operation, thermal expansion of the ball screw is
occasionally approximated using a model in the control. Because the temperature profile is difficult to measure during operation and is infl uenced by numerous factors – such as the wear of the recirculating ball nut, the feed rate, the cutting forces, the traverse range used, etc. – considerable residual errors up to 50 μm/m can occur when this method is used.

The ball screw is sometimes provided with fixed bearings at both ends in order to increase the rigidity of the drive mechanics. But even very rigidly designed bearings cannot prevent expansion caused by local heat generation. The resulting forces are considerable. They deform the most rigid bearing confi gurations and can even cause structural distortions in the machine geometry. Mechanical tension also changes the friction behavior of the drive, thus
adversely affecting the contouring accuracy of the machine.

Due to these restrictions, the drive accuracy that can be attained by taking the described additional measures cannot be compared with closed-loop operation using linear encoders. Also, the additional measures for semiclosed-loop operation cannot compensate the effects of changes in the bearing preload due to wear, or elastic deformations of the drive mechanics."

And, we're talking about milling machines, not "diamond turning hardware", so that point is totally irrelevent. But I'm guessing whatever that machine is, it is probably not moving hundreds of pounds of cast iron around at hundreds of inches per minute for hours at a time.

The *type* of encoder is irrelevent, but for practical reasons, these days nobody is going to build a new machine using tachs, resolvers and other such stone-age components. Modern control systems are, for all practical purposes, all digital.

Haveing rerouted a few air conditioning ducts in my time I can't dismiss the relaity of thermal expansion. However a machine will reach an equilibrium at some point.

If you happen to be running a VERY consistent cycle in a temperature-controlled environment, that can, MAYBE, sometimes be true. But, in general, it's simply not, as a little research should quickly prove to you. Heat will be generated where there is moving contact between parts - bearings, ballscrews, ways, etc. If, as is VERY common, you spend long periods of time working over a small range of locations, those areas will heat up, and others will cool down. If you now move to a different position, you're working with cooler components - FOR A WHILE. And it will virtually NEVER be the case that the entire machine is at a constant temperature for any length of time, which means thermal expansion WILL affect precision. And given that each component will expand and contract at it's own rate, based on it's own temperature, it's a VERY complex problem.

Again, if your position were valid, why do virtually all modern VMCs use direct positional feedback to compensate for thermal errors? Surely they could save a great deal of money by leaving all that out?

Regards,
Ray L.

[waltpermenter]

hey ray, other than your x2 that's been collecting dust since 08 when you got your chinese kneemill, do you have any experience actually using another benchtop machine?
How about any real world machining experience using industrial equipment? Ever hold a job in the industry? just asking.
walt

[Padrino]

Please let us know which ball screw you end up purchasing.
Thanks

[bigspike]

Please let us know which ball screw you end up purchasing.
Thanks

I will and hopefully report back after installation and testing.

To add another aspect... I had decided on 3/4" ball screws for X Y & Z, but was interested in opinions regarding using 5/8" for the X & Y

ALSO ... lets get real with the specs here. I will be ecstatic if I can get repeatable output that is square, round, and within .003"

My primary concern is with non quality rated Roton ball screws vs Chinese ??? rated vs precision rolled ball screws that I may not be able to afford.

Roton doesn't give any rating to their ball screws and their website lists applications of :
hospital beds
medical examination equipment
servo drives
log splitters
dish antenna drives
door actuators
scanners and test equipment
plotters
robotics
web tensioners
automatic pilots
trim tab drives
back gage positioners
fork lift remote accessory drives

Nothing about CNC machine tools. I certainly hope to achieve better accuracy than is required by a log splitter!

[SCzEngrgGroup]

For an RF 45, 3/4" screws should be fine. You might consider 1" for Z only. The Chinese ones from seller linearmotionbearings2008 on E-Bay are much better than the Rotons. If you want better still, but somewhat more expensive, look at Nook XPRs, which have a rated lead error or +/-0.001"/foot. You can use the XPR screws with less-expensive non-XPR nuts, and double-nut them with spring pre-load to get near the best of both worlds with very reasonable cost.

Regards,
Ray L.

[Padrino]

I am looking at building a cnc machine perhaps similar to a xzero machine maybe since the raptors are no longer available in the size i am looking for..

Could you perhaps post a link of this roton ball screw. I am not familiar with them. I was looking at the SBC ball screws. I am thinking of using their SBI linear bearings as they offer a precise travel with similar specs to THK or thompson but at a fraction of the prices. But just want to see what others are choosing.

Thanks in advance.
Rob

[SCzEngrgGroup]

I am looking at building a cnc machine perhaps similar to a xzero machine maybe since the raptors are no longer available in the size i am looking for..

Could you perhaps post a link of this roton ball screw. I am not familiar with them. I was looking at the SBC ball screws. I am thinking of using their SBI linear bearings as they offer a precise travel with similar specs to THK or thompson but at a fraction of the prices. But just want to see what others are choosing.

Thanks in advance.
Rob

Download the ballscrew catalog in PDF from the Nook website, and there's a whole section on XPR.

Regards,
Ray L.

[691175002]

I'm four holes away from being able to map my ebay screws so I'll try to get some numbers this week.



I'm also pretty interested in seeing the numbers. Any bets?

[rwskinner]

Walt, It's time to lay off and stop starting flame wars.

hey ray, other than your x2 that's been collecting dust since 08 when you got your chinese kneemill, do you have any experience actually using another benchtop machine?
How about any real world machining experience using industrial equipment? Ever hold a job in the industry? just asking.
walt

[waltpermenter]

Walt, It's time to lay off and stop starting flame wars.

excuse me but mind your own beeswax, i was asking simple questions of ray, you are the one trying to instigate something. If he's not going to answer I'll just assume the answers are no, no big deal, no insults, just asking
walt

[Padrino]

Download the ballscrew catalog in PDF from the Nook website, and there's a whole section on XPR.

Regards,
Ray L.

Ray and others,
I am initially looking to build a machine with a 32x 48" travel and was considering using a 2525 precision rolled ballscrew for the X axis however, nook doesn't have one in the XPR accuracy. Perhaps I should settle on one with a smaller lead. What are your thoughts on this?

Cheers