Design Critique for RF-25 Ballscrew Upgrade

I have a Rutland RF-25 mill/drill that I have been planning on upgrading to a full CNC for a while, and I currently have a "preliminary" CNC implemented using Paul Thompson's CNC kit. This uses Nema 23 motors, and actuates the original ACME lead screws by turning the pawls via a set of gears. This works fine enough for now, although it is very slow, and the fact that it uses the original lead screws means that there is about ~15 thou of backlash on X and Y axes that can't be reasonably mitigated. As such, I decided to use this preliminary CNC setup to design/build a much more robust/precise CNC, using larger motors in a direct-drive configuration with ballscrews with double ballnuts. Right now I am only upgrading the X/Y-axes, the Z will come later. The ultimate goal is to control the entire system using LinuxCNC.

I spent a very long time preparing for this upgrade -- I dissected, measured, and modeled my machine along with all its gloroius imperfections/oddities. I filled the base/column with epoxy granite along with 80 lbs of lead shot, lapped all of the ways, repacked my spindle about half-a-dozen times before I was satisfied that the runout was under control, and spent entirely way too much time tramming the column. I'll post a thread on all this (+ a bunch of other upgrades) I've done over the years at some point, but until then it will suffice to say that this machine is now very tight indeed. The only thing I haven't yet figured out how to do is minimize the clearance between the splines of the spindle and spindle sleave, as this causes chatter occasionally and drives me bananas.

I am largely modeling my ballscrew upgrade after KCJ's supurbly documented upgrade. I would love it if I could mount the ballscrew supports under the table like he did for compact awesomeness, but I decided not to on account of the fact that I only have one CNC machine available to me, and making the modifications would be too much of a schlep to reasonably overcome.

I'm not particularly looking for any specific help, but I would appreciate it if I had some folks take a close look at my design. I'm an amateur engineer (at best), and I would love it if some folks who are actually knowledgeable about this sort of thing would knock me down a peg or two. So far I only have the motors, nothing else has been ordered.

I've attached some renders from my design in this thread, but there are plenty more on my google drive. I've also included the CAD files for everything in Solidworks 2020 format, in case anybody has the newest version. I haven't converted it to any universal format (STEP, etc) because I don't see the value, however if anybody thinks it's helpful I can provide this. Also, I color coded the parts/renders to indicate custom made parts by me (orange) and yellow for the lead screws.


MOTORS
As mentioned, I already have the steppers, and they are hybrid closed-loop steppers from China/Alibaba (Nema34, Sutai M/N: ST86HSS801). These are the same style that KCJ uses in his machine, and are similar to Leadshine models. These motors are, obviously, oversized for my machine, and I do not plan on hitting high speeds. My decision to go with this motor was primarily driven by the fact that for being a pretty substantial motor, the inductance is quite low (3.5 mH). Even though I will be using this motor to direct-drive my ballscrews, with microstepping I will be able to obtain a resolution that more than exceeds my requirements. Another added benefit to using these particular motors to direct drive the ballscrews is that I will always be in the linear part of the torque-curve, since the motors will be driven at a slow speed, at least relative to their full capacity. Here are the specs:
Formfactor: NEMA 34
Shaft: 14mm diameter, 20mm length
Inductance: 3.5 mH
Current: 5 A
Resistance: 0.43
Holding Torque: 4.5 N-M


BALLSCREWS
I decided to go with 1605 ballscrews from Chai. To be clear, this was not a financially motivated decision, it just seemed like this was the easiest. I've read the threads about all the horror stories regarding the quality, but at the same time there is an abundance of people who have had great experiences. At the end of the day I'm hoping to hold my machine to about 0.001" anyway, and I didn't see the need to go with ground ballscrews. I care a lot more that the backlash is going to be consistent than as close to zero as is physically possible, as it is my understandingthat LinuxCNC has superior backlash-compensation when compared to Mach3. I'll also be using a double-ballnut configuration (more specifically, 2x single ballnuts per axis) with disc springs (aka Bellevilles) to provide preload. The ones that I chose are the only ones that McMaster seemed to have that physically fit my design, however I'm quite confident that they will provide too much preload (flat load is 530 lbs each). I'll look elsewhere for some different options, but obviously I will have to determine the best ones (and what compression) empirically. I'm also open to suggestions :)

The drawings for the ball screws for the X/Y-axes are attached to this thread. The end machining was lifted from the documentation for the BK/BF-12 supports, I'll make sure I add tolerences before I send it over to Chai. Also, what kinds of chamfers do yall recommend, and where? Another question I had for the broader audience is whether anybody bothers to have the ball screw threads for X/Y operate in opposite parity (the threads in the original X/Y screws go in opposite directions). This is obviously something that the controller can switch instantaneously through a software setting, I'm just curious.

MOTOR COUPLERS
My design uses Misumi CPLCN32-10-14, which allow for all three types of misalignment and have "zero-backlash." I didn't put much thought into this to be honest, and I'm happy to hear suggestions.

BALLSCREW SUPPORTS
I will be using the BK-12 and BF-12 supports. I played around with using the flange-mounted versions of these, as I wanted to bring the center of mass as far inward as possible, however I couldn't make it work any better than using the BK/BF. The mounts for these are going to be the same width as the supports themselves (aside from the BF, which I did on purpose to add additional mass on the opposing side for balance), however when I tap the holes for mounting them together I will make sure there is a bit of clearance between the support and the machine mounting surface so that the system will not become overconstrained. For the moment I am planning on only attaching the mounts to the machine using the 2x original 5/16-18 threads, but I can drill/tap additional holes if it becomes necessary.

I'm thinking of sourcing these from SYK, as they seem to be one of the few vendors who are honest about what grades of bearings they have, and also the option to upgrade. They also aren't particularly cheap (or expensive), which leads me to believe that the quality may be well-suited for my intended application. Aside from these things, my decision to go with them is pretty arbitrary, and again I would love for some additional input for some alternative vendors. The CAD models uses solid models from SYK that I downloaded from their website, as were the specifications on the end machining (but as far as I can tell, for BK/BF-12 these values are pretty standard).

MOTOR MOUNTING
I had to convince myself not to design/machine a monolithic motor/support mount that bolted directly to my table, as it was too much of a hassle and I couldn't convince myself that the gain in rigidity would have translated to anything measurable. I found some steel female-threaded standoffs with 5/16-18 threads (1.75" long) that seem like they will work well. You'll also notice that I used 3x of these on the X-axis motor and 4x on the Y-axis. I prefer 3 mounting points whenever possible so as to not over-constrain the system, and while I was able to accomodate this for the X I was not able to for the Y, at least not to my liking. The three legs for the X design form an isosoles triangle, with the center aligned with the axis of rotation of the lead screw, but I couldn't make it reasonably work for the Y without moving the motor down a significant amount. For this reason I just went with 4 standoffs, which is fine probably.

MATERIAL SELECTION FOR CUSTOM PARTS
I am considering making all of the custom mounts out of Mic-6 aluminum tooling plate. I have a surplus of this material, and it is an absolute dream to machine, and has fantastic dimensional stability. However, it is not as strong as 6061, and for this reason I am not sure whether it is the best choice. I imagine I'm probably splitting hairs here, and if anyone has any feedback on this specifically I would love to hear it. I do not currently have any sort of coolant on my machine yet, and even though I have tons of Carbide end mills, my motors are weak. I have my maximum speed @ 10IPM with aluminum (28IPM is true unloaded maximum, but set to 10IPM so I don't have to worry about missing steps), and it would probably be half that for steel. I also have plenty of Mic-6 and possibly enough 6061 on hand, and have very little steel.

That's all folks! Let me know what you think :)

Cheers,
IHateMayonnaise



https://www.cnczone.com/forums/attac...d=456214&stc=1



https://www.cnczone.com/forums/attac...d=456216&stc=1

That looks like a lot of CAD modelling,which I suspect you will be glad to have available as the project progresses.I have two comments and they are pretty superficial;why two ballnuts?The standard ballscrews will probably have a vast amount less backlash than you are used to.Second,How do you propose to move the Z axis?Will you be using the manual height adjustment for coarse adjustment and then adding stepper drive to the quill feed?

I think you will be very pleased with LinuxCNC,I know I am but I use it for a router with less demanding levels of accuracy than you are seeking.It is a very capable piece of software at an unbeatable price.

Quote:

---

Originally Posted by routalot

That looks like a lot of CAD modelling,which I suspect you will be glad to have available as the project progresses.I have two comments and they are pretty superficial;why two ballnuts?The standard ballscrews will probably have a vast amount less backlash than you are used to.Second,How do you propose to move the Z axis?Will you be using the manual height adjustment for coarse adjustment and then adding stepper drive to the quill feed?

I think you will be very pleased with LinuxCNC,I know I am but I use it for a router with less demanding levels of accuracy than you are seeking.It is a very capable piece of software at an unbeatable price.

---

A lot of modeling indeed. My primary motivation for modeling my machine so thoroughly was to learn Solidworks/CAD modeling in general. I had never used it before starting it, and at the end I'm very familiar.

My reasoning for going with double ballnuts is to reduce the total backlash, and also the fact that it is minimal effort to do two ballnuts rather than one. If backlash compensation works well in LinuxCNC (allows for anti-symmetric compensation, etc), then I'd say two is not necessary, but for now I'm planning on double.

My z-axis is currently CNC'd using Paul Thompson's kit, which simply turns the original worm gear (see attached). This works pretty well in the z, it's plenty fast and has only ~1.5 thou of backlash. I've never run into the Z being underpowered anyway, so I reason that it is fine for now. I'll eventually do a ballscrew (I already have the motor for it).

As a side note, I'm very eager to start using LinuxCNC, however I am driving myself crazy trying to source components for the machine. A common thread that I've noticed is that latency/jitter timings are hardware related, and they should be paid attention to in order to have a responsive system with the real-time kernel. I just can't find any recent motherboard/cpu/memory combos listed anywhere which are known to give good results, and I'm not particularly interested in buying a bunch of hardware just to optimize the latency...

Loads of information there...

I have seen information on those stepper couplers failing, they seem to fatigue out. Also, .0015" backlash isn't bad, certainly not for a machine of this design and size. You can halve that or more with double nuts. The big issue with these machines isn't necessarily reducing backlash, I think anything under .002" is more than adequate (I don't own a RF25 but rather a G0704). The issue is getting the gibs tight enough to keep everything tight without stalling steppers. You can easily produce more than .002" of play in the table by reducing gib tightness for the sake of preventing lost steps.

Quote:

---

Originally Posted by CL_MotoTech

Loads of information there...

I have seen information on those stepper couplers failing, they seem to fatigue out. Also, .0015" backlash isn't bad, certainly not for a machine of this design and size. You can halve that or more with double nuts. The big issue with these machines isn't necessarily reducing backlash, I think anything under .002" is more than adequate (I don't own a RF25 but rather a G0704). The issue is getting the gibs tight enough to keep everything tight without stalling steppers. You can easily produce more than .002" of play in the table by reducing gib tightness for the sake of preventing lost steps.

---

Did I say .0015"? I totally meant 0.015".

I agree, there are many sources of backlash in the gibs. I minimized this best I could by lapping the ways/gibs with Timesavers lapping compound, and minimizing play between the original gib retention screw and the gib notch when the direction is reversed (essentially minimizing the gib sliding). I also shimmed the thrust bearings to minimize play there as well. I also replaced the ACME table nuts for both the X and Y, which helped a lot. The problem I have is mainly that the motors are not able to provide the optimal torque to the screws without stalling -- right now the gibs as well as the compression screw on the lead screw nuts are a lot looser than I would prefer.

I wouldn't think latency will be any kind of issue with almost any computer from the last ten years or so.Particularly if it does nothing more than controlling the one machine.Even with fairly fine microstepping there can't be that many (micro)steps to travel the entire length of any of the axes.Is there any kind of used computer available for running a latency test?You might be pleasantly surprised.

With the revised target of 0.015" I would say that doubling the number of ballscrew nuts isn't likely to be essential as they are likely to be much better than this when they leave the factory.In three or four years-that may change.One thing to keep very firmly in mind is that with those big NEMA 34's turning a 1605 ballscrew you will be operating a setup that can produce very large forces.I did once have a phone app that did the calculation for motor torque and thread pitch and it told me my NEMA 23's could exert a force of 800lbs.I can stall my router with a lot less force than that and I wonder where the weak link will be in this project;not trying to depress you,but to steer you in the direction of keeping a watchful eye on the most likely area to cause trouble.Maybe that would be a good incentive to experiment with FEA within Solid Works since a very thorough model already exists.

He said 15 thou, or .015", I just converted incorrectly in my brain. I'm kind of dense like that.

That is indeed terrible backlash.