[guru_florida]

Here is some pictures of my progress on my homebuilt cnc machine. My goals are for a precision/accurate machine for small robotic parts, pcb's, etc.

I am trying to decide how I am going to secure the fixed-end bearing (dual-row thrust/axial bearing) to the ballscrew. Unfortunately, there is no thread on either end to screw a nut that will *squeeze* the bearing. I have a good interference fit already. I chilled the ballscrew and heated the bearing to fit the bearing on and still had to bang it on. I probably couldnt bang it back off now that its cooled...but is this enough strength? (probably not.) You can see the bearing I am talking about in the pic. I have a closeup too.

btw: you can see more detailed pictures of my machine so far at:
http://photos.colinmackenzie.net:81/cnc/

Thanks,
Colin

[RICHARD ZASTROW]

Colin, If you "banged" on the bearing(s), odds are they're already damaged. Is it possible to have threads installed outside of the bearing seat area? I don't know what your "dual-row thrust/axial bearing" type of construction is or what level of precision/accuracy you require. Personally, I'd opt for 60 degree angular contact (AKA ballscrew support) bearings for accuracy and axial support. As NC Cams explained to me some time back, double row angular contact bearings have internal clearances and therefore some end play. I do believe there are some with seperate inner races that can be preloaded but I've never used them as yet.

Good luck

Dick Z

[guru_florida]

Maybe I shouldnt have said banged. The bearing still runs very smoothly. I only tapped on the inner race with a some pliers.

How can I apply some threads to the shaft? I have taps, but this would be the opposite. Have to get a metal shop to do it?

I am not sure the angle of the dual bearing. It doesnt seem to have any play at all. It seems a good quality. It's a high speed bearing if that means anything.

I am still learning a lot about designing cnc machines. Learning cnc is like peeling away an onion, each layer exposes more details/issues!

Colin

[RICHARD ZASTROW]

Colin, Looks like you got a little ahead of yourself. lol Seriously, you might want to stop and plan/design ahead. I have no idea of where you are in preparing your next move or what is already completed. Your bearing must be mounted in some sort of support; the nut also; are you going to support the opposite end? if so, with a "floating" bearing?

ref. metal shop to do it? probably, choose a good one. Assumptions are a bad idea but here we go. If your shaft end is case hardened (assumption 1) it may be a simple grind thru the case hardening, turn and thread for a bearing locknut. May require a spacer/adapter to clamp the bearing against a shoulder. or?

There's a lot of good info here. Peruse the project logs and other forums, most of youre future questions have already been answered. If you don't find it, ask specific questions and you'll get more answers than you need (or even want).

Another load of info is at the websites of the bearing, ballscrew and linear way systems companies, especially the engineering and technical support areas. They'll give you all the shaft shoulder and housing shoulder dimensions etc. Read them, good stuff.

Hope that helps for now, adios

Dick Z

[guru_florida]

I am designing the entire thing in solidworks. Here are some pics from SW for the fixed end of the bearing mount. (no threads or screws are shown in the SW models.) The mount is made from two pieces of rectangular aluminum.

Yes, the other end is floating too. The ballnut carriage mount is done and is what is holding the ballscrew in place in the pictures. I am relieved that it lines up with the two end holes perfectly (where the bearing mounts go).

I am getting a lot of design ideas from 5bears.com (my overall inspiration) and what I have seen from people on here. I am a competent engineer but I lack a lot of the proper tools.

C

[RICHARD ZASTROW]

Cory, Obviously you're further on than I "ASSumed". My bad. Judging from your latest post, you're using direct coupling to the axis motor. It appears you have included an opening to access the coupling. That looks more sturdy than 4 round rod/spacers and supports the bearing as well.

Are there any labels on your bearing? Maybe one of us can identify it for you.

What tolerances are you requiring? I tend to go for extremes. That often leads to expensive over specifying components.

Dick Z

[djr76]

With as little as that machine is, you can just make that housing 1 pc. It would be way more accurate.

[guru_florida]

Oh yes, the 4 round spacers was definately a "temp" thing! I will be using flex couplers to attach the motor and screw shaft, but yes, direct drive. The top hole is to access the flex coupler.

I wish I could make this out of one piece right now, but I have very limited tools. A dremel tool, portable drill and access to a drill press 40mins away! I am suprised things have worked out this well so far! I would buy some tools but I live in a small apt. My neighbors are going to hate me when I finish this machine!

I am designing my machine as though I will get .0001 precision. And yet, only hoping I get anywhere near that!!! (few tools and all.) I am counting on the systemic process of tightening the bolts to makup a lot of the accuracy. There is some clearance in most of my bolt holes. The bearing holes I cut small and trim with my dremel for a tight fit.

I am sure I will end up with a machine that can make some nice robotic parts and do a lot of the things I want. In the process I will learn a lot more about cnc, the pitfalls and traps, and continue to upgrade the machine in the process...using the machine to create itself - so to speak. Going back to the onion analogy, I think making your first machine is a right of passage...then you truly understand the problems. Here I am making a machine that I have never actually seen in real life much less used before! (Well, once I got a tour of the cnc lab at my university 10 years ago.)

Colin

[guru_florida]

Here is the link to the bearing on ebay:
http://cgi.ebay.com/ws/eBayISAPI.dll...e=STRK:MEWN:IT

[RICHARD ZASTROW]

That link indicates a standard (normal axial clearance value .0002"-.0008") double-row angular contact bearing with (2) shields. (not seals). Do you have any other markings (C2 C3 or?) on the bearing? Any on the screw or nut? We might be able to determine tolerance stack potential before assembly hopefully avoiding obvious excess. Actual total compliance will have to be measured after assembly, as there are many possible contributors to "problems".

Dick Z

[RICHARD ZASTROW]

Colin, Ref. your above post (#8) by flex coupling I hope you mean bellows or similar. The type of flex coupling that permit radial "torqueing" will introduce lost motion between the axis motor (and encoder/resolver) and the ballscrew. Not good for .0001" stuff. The bellows type coupling compensates for limited angular and axial misalignment without changing rotational displacement, as long as appropriately sized.

Dick Z

[guru_florida]

The flex couplers are these:
http://cgi.ebay.com/ws/eBayISAPI.dll...MEWN:IT&ih=011

I think they will work without contributing to error. As for the bearing. I played around with it more and it probably isnt the greatest. I think I can detect a little play, mostly by applying a moment to the bearing perpendicular to the rotation axis. Its not visually apparent, but I can feel it. It will have to be replaced later I'm sure...along with a lot of the other parts.

I really appreciate your time and help so far...I am learning with your feedback.

Colin

[RICHARD ZASTROW]

Colin, The couplings in your post #12 will absorb shock, compensate for minor misalignment etc. They will also contribute to lost motion. They may work fine for a router but you said you wanted to design for .0001" tol. in small metal parts.

We've identified .0001-.0008" in bearing end play. The coupling will contribute (minutely) more "slop" and we have not yet determined accuracy of screw/nut. This is only one axis. What happens when we accumulate the error potential of 3 or more axis?

If you really want to hold .0001"'s tolerances you must plan for it. There's always the "law of compensating variables". I'd bet on Murphy's Law, "whatever can go wrong will".

BTW, did you ever find i.d. info for the screws/nuts. We might be able determine the error potential there as well. C0-C7 ?

Keep working at it.

Dick Z

[guru_florida]

I was just rereading this thread to go over some of what Richard was saying again. I realized I typed .0001 tolerance when what I meant was a thousands'th of an inch, 1mil or 0.001! Thats a big difference when you are talking that small a tolerance!

It's been a long wait on ebay for the right ballscrews but I finally have a set of C3 ball screws coming with the bearings already present. These should be what I am looking for.

Richard: If you are still around, what makes those flex couplers I got susceptible to lost motion? I assume its the flex part will compress under load? I thought those were metal couplers when they turned out to be cheap plastic ones! Sometimes you get burned with ebay. I should have been smarter there.

I saw a metal flex coupler on another ball screw I got for the z-axis. The idiot who took it out didnt take the time to remove it and instead just ripped the motor off the coupler so it was damaged, but the idea was it was two flat plates and some thin ribbon like metal strips attaching the plates and the motor end and the screw end. I can see how it would flex in the misalignment axis while staying rigid in the rotational axis.

[RICHARD ZASTROW]

Colin, Yes, I'm still here, the grim reaper hasn't got me yet.

Even to attain a .001" level, you must consider the error accumulation of all components.

The flexible coupling will compress to some degree. The more force applied the more it will compress. Granted it may be a minuscule amount and acceptable.

How much lead error and backlash are in the leadscrew/nut combination? Again, it may be an acceptable level.

Is the geometry of the machine in question in alignment within acceptable error limits?

Add these together and Murphy's Law say it will all go wrong at the same time. Fortunately, there is also the Law of Compensating Variables where the errors will cancel each other out. LOL

The equipment I work with is unforgiving and requires extremes in accurate components so I tend to be over critical. Sorry 'bout that.

I'd still recommend the metal bellows type coupling. Actually, a good belt drive might be better than a direct coupling. I've had bad experience with direct coupling with very small misalignments causing motor bearing failures and encoder problems from vibration transmitted thru the servomotors. Problem went away when direct coupling was replaced with a belt drive.

Good Luck

Dick Z

[Pres]

I'll 2nd the Dick Z observations.

Belt drive has a number of advantages: backlash, easier inertia/resonance mods, can be more compact, and adjustable ratios.

The attainment of 0.001" on home machines is attainable, but difficult and tedious.

Getting 0.0001" borders on ridiculous and is, for all practical purpses, unattainable on home machines.

By accident, maybe on one surface of one part.

Nice to think about achieving it however.

Pres

[guru_florida]

Even to attain a .001" level, you must consider the error accumulation of all components.

Thats what I am concerned with now. Actually making the numbers work. I have the table working now with temporary parts and I have spent some time learning different software tool chains. Everything from design cad, controller software, conversion utilities, etc, and using this to cut some things such as isolation routing a pcb board, some wood cuts and very light metal. It has worked great so far, even with a bad temp Z-axis. Now that I've actually seen cnc in action, I feel like I can begin to understand the design caveats. So I am back to the drawing board to complete a rigid z-axis and complete the base.

The flexible coupling will compress to some degree. The more force applied the more it will compress. Granted it may be a minuscule amount and acceptable.

I will look into this more.

How much lead error and backlash are in the leadscrew/nut combination? Again, it may be an acceptable level.

This is the problem with ebay, hard to get specs. I just about freaked right now. I was checking the part number on my NSK JAPAN 92A4 W1406FS - 1 - C3T's that I have on order and never noticed the T typically meaning Transport. I must have gotten caught up in the excitement with finding a set of rare C3's on ebay (damn). A google search finds an axial play of 0.005. Luckily this is mm not inches so its <0.0002.

I dont know the bearing part# but being C3 I hope the ball bearing they already contain are from the equip and likely thus the right match.

Is the geometry of the machine in question in alignment within acceptable error limits?

It appears to be in good alignment now and I've put little to no effort into proper assembly alignment, must be that Law of Compensating Variables working for me, lol. I have however put a lot of work into the cad design and the assembly mates. I use SolidWorks and it complains if mating doesnt match with even the smallest misalignment. My SolidWorks model rebuilds well even with a deep change. What I mean to say is I think my patience with the cad design has helped my final assembly in keeping parts in agreement. (kudos to cad.)

I have oversized certain holes appropriately to take up small misalignment with the ballscrew/nut combo mounts. Yes, I'm counting on the Law of Compensating Variables here but I am anticipating a little Murphys law since some of my parts are used and from ebay! During final assembly I will use the micrometer and a pre-defined order of assembly ensure alignment.

Am I on the right track here?

Add these together and Murphy's Law say it will all go wrong at the same time. Fortunately, there is also the Law of Compensating Variables where the errors will cancel each other out. LOL

I initially thought Law of Compensating Variables was a joke, though I understood the concept. Google can only find this reference, "...Makes just about as much sense as the law of compensating variables, aka: it all works out in the end." (it also finds this thread!)

The equipment I work with is unforgiving and requires extremes in accurate components so I tend to be over critical. Sorry 'bout that.

Thats why I liked this thread, you're a man in the know. Please dont apologize.

I'd still recommend the metal bellows type coupling. Actually, a good belt drive might be better than a direct coupling. I've had bad experience with direct coupling with very small misalignments causing motor bearing failures and encoder problems from vibration transmitted thru the servomotors. Problem went away when direct coupling was replaced with a belt drive.

Though I knew belt drives were good, I didn't know they could compete with direct drive with proper couplers. I figured I would upgrade to servos instead of steppers in the future, perhaps I will switch to belts then.

Thank you,
Colin

[RICHARD ZASTROW]

Colin, Keep trying for .001". Realistically, as Pres stated, a homebuilt DIY machine made with extrusions and without the proper equipment is not going to get you there. You should do your best but don't be discouraged if you can't reach that target.

It's like that old adage, "speed costs money. How fast do you want to go?". In this case it's more like "accuracy and precision are expensive and time consuming. How much time and money do you want to spend?"

One of my "cohorts in crime" and I undertook a project. (his idea) This was a (6) axis CNC gear hobbing machine. The first machine took 3 years to design; produce and purchase proper components; build, test and tweak the first machine. It was and still is more accurate and faster than any machine of it's type. A number were built and are all in one place successfully producing gears commercially. Of course, it took twice as long and cost twice as much to produce as originally planned. I'm working with him again on other projects and loving it. LOL

Keep on truckin'

Dick Z

[guru_florida]

Hey, Im pretty excited with what I can do with it now and Im at about 1mm accuracy!!! hehe. (due to z-axis, not the table.) Even 10mil would be good for 90% of what I want to do --- mostly robotics, prototype pcbs, and perhaps a few moulds for custom electronic boxes.