Grizzly G0704 Cnc Conversion

I'm chasing about .007" in the X axis. Haven't checked Y yet. Here's the link to my thread. It's been a challenge. http://www.hobby-machinist.com/threads/chasing-backlash.55182/#post-456709

Tom S.

Wow. Just read that whole thread. Not sure I understand all of it, but most of it is just like the hardware on the G0704 conversion. I'm a lot more concerned about mils here and there than I was before.

I note one of the guys said "0 is acceptable backlash and with double ball nuts should be achievable". Zero? To how many decimal places? I wonder if that's realistic. If you took a production mill from someone like Tormach and measured it, would you really get 0.000? If you do get 0.000, do you have to tweak that periodically to keep the low backlash? Parts wear out, after all.

It's not a great comparison, but my A2ZCNC/Sherline system has backlash in the .005 range.


Bob
 
Wow. Just read that whole thread. Not sure I understand all of it, but most of it is just like the hardware on the G0704 conversion. I'm a lot more concerned about mils here and there than I was before.

I note one of the guys said "0 is acceptable backlash and with double ball nuts should be achievable". Zero? To how many decimal places? I wonder if that's realistic. If you took a production mill from someone like Tormach and measured it, would you really get 0.000? If you do get 0.000, do you have to tweak that periodically to keep the low backlash? Parts wear out, after all.

It's not a great comparison, but my A2ZCNC/Sherline system has backlash in the .005 range.


Bob

I'm relatively new to CNC, about 2 years experience, but parts do wear in and need periodic checking and adjustment. This is particularly true when you are wanting backlash to be in the .000" to .001" range. As an example when I first installed the double ball nuts I was able to machine a circular pocket that was round to within .001" - .002". Now that I've run in the ball nuts for a few months I've got excessive backlash and therefore out-of-round bores. Now that the ball nuts have run time on them and get the backlash to an acceptable level it will stay that way for a while.

My goal is zero backlash but I don't know if I will get there. I'm trying though.

Tom S.
 
I note one of the guys said "0 is acceptable backlash and with double ball nuts should be achievable". Zero? To how many decimal places? I wonder if that's realistic. If you took a production mill from someone like Tormach and measured it, would you really get 0.000? If you do get 0.000, do you have to tweak that periodically to keep the low backlash? Parts wear out, after all.

That was me. With preload on the ball nuts there is an interference fit, so therefore no axial movement is possible between the ballscrew and the ballnut. This also assumes that the ball track is not somehow floating inside of the ballnut, not sure if that is even possible. This also applies to the preloaded end support bearing, if everything is tight no axial movement is possible. Now, having said that, the overall system may have flex, stretch, or looseness in it that would manifest as backlash.

I would expect the Tormach right off of the assembly line to have no backlash if it was setup correctly. After 30 days of run time I would expect it to loosen up a bit and would expect to need to make adjustments. Once that is done, periodic adjustment might be necessary but at a much lower frequency to keep it in proper tune.
 
That was me. With preload on the ball nuts there is an interference fit, so therefore no axial movement is possible between the ballscrew and the ballnut. This also assumes that the ball track is not somehow floating inside of the ballnut, not sure if that is even possible. This also applies to the preloaded end support bearing, if everything is tight no axial movement is possible. Now, having said that, the overall system may have flex, stretch, or looseness in it that would manifest as backlash.

I would expect the Tormach right off of the assembly line to have no backlash if it was setup correctly. After 30 days of run time I would expect it to loosen up a bit and would expect to need to make adjustments. Once that is done, periodic adjustment might be necessary but at a much lower frequency to keep it in proper tune.

If that came across as harsh or critical, I didn't mean it that way. By background, I'm an engineer (electrical - so mechanical is mostly out of my "comfort zone") and I'm just trying to understand what's going on. Obviously, nothing is zero backlash to the millionth of an inch, and probably not to a tenth (.0001"). It seems like something is always going to show up that looks like backlash. If it's always there, when do we stop chasing things? When is it "good enough"? I guess we all make that call.

As I said, I have backlash issues on my newly converted G0704, too. Quite a bit worse than Tom's (.010"). Before I take apart my mill, I'd like to have a good idea of what I'm looking for.

I had to re-ball one of the ballnuts off of one of Chai's ballscrew/nut combinations and while I know there's not supposed to be axial motion, I didn't see anything in there that would guarantee that. I didn't see a shim of any kind. I didn't take the two halves apart, though, so it's possible one was in there.

One of those things I've read is that people put larger ball bearings into the ballnuts and take out backlash that way. The way I think of it, there are something like 8 races in the linearmotion ballnuts. If you have 5 mils of backlash, maybe that means you make the balls .005/8 bigger or .000625". The ones I just bought were .1244", this would say to use 0.1250 bearings. Turns out the ebay seller I bought from sells them. Is that .005 divided over the 8 races the right way to think of this?

The thing that gets me is that ballscrews (really all screws) have imperfections from turn to turn so that the exact distance between turns wanders around the ideal. When I was looking at my screws with a dial indicator, some spots seemed to be off about .001" from one tenth (0.100") to another but averaged out to be right. It worked out that going from tenth to tenth it could be 0.1.00, 0.099 or .101, but over the few .500" intervals I measured it was perfect. Regardless of lost motion in the machine, that .001" is going to show up on some cuts. There are different "classes" of screw that have tighter or loser tolerances on this, and as you'd expect, the higher the class of screw, the more expensive. Now, having a circle out of round by .001, or a hole location off by .001 isn't always going to matter, but it's always there.

What about the way they deform under load? Under load, the balls go out of round and the thread they're in changes shape, too. It probably doesn't contribute to backlash, but it contributes to overall accuracy.

Sorry if this is rambling. As I say, I'm trying to understand all this. Is there a general tutorial approach to tracking down where backlash is coming from? A "do this, check that"? kind of thing?
 
With the stepper motor removed set up a dial indicator on the end of the screw, lock the table, then rotate the screw by hand. If the screw is moving in and out your problem is in the axial bearings. Thats where mine was. If it does not move it could be in the ball screw, but .010" is an awful lot...makes me think the balls might not be in their proper tracks or something???
 
The .010 is the one I re-balled. The one I never touched is .012. I figure that's "in the noise". It's got to be the axial bearings.

On the topic of running the table longer distances and verifying it moves as commanded, I was just measuring the length of 2 1-2-3 blocks I have. I have three sets of calipers. Two of the three read 6.000 (and they all throw in that meaningless fourth decimal place that's either 0 or 5). One set read 5.998. I think that one will be relegated to the woodworking side of my shop.

I assume it's obvious that I wouldn't have built it up if I knew there was that much backlash.

Edit to add: left out a piece of the first sentence, second paragraph.
 
A special thanks to jbolt for enlightening me on this test. You can isolate the source of your backlash by mounting a dial indicator to something other than the table, e.g. the column or the saddle then set the dial indicator tip to the end of your table. Remove the stepper motor and the AC bearings. Push the table in one direction and set the DI to zero. Now push the table in the opposite direction and note the DI reading. It should be zero but if it isn't then you have clearance in the ball screw/ball nut assembly. This is a test I have not done yet but intend to do when I get back in the shop on Tuesday.

I have done the test Dave described in his post above and I have zero screw movement in relation to table movement with the AC bearings assembled and preloaded. This is telling me my backlash problem is not with the bearings. Hopefully I will find the source of backlash is in the ball nut.

Tom S.
 
With the stepper motor removed set up a dial indicator on the end of the screw, lock the table, then rotate the screw by hand. If the screw is moving in and out your problem is in the axial bearings. Thats where mine was. If it does not move it could be in the ball screw, but .010" is an awful lot...makes me think the balls might not be in their proper tracks or something???

It makes sense to me that if I can turn the screw without turning the motor, I need to crank that out first. I never did that because I didn't have a good way to do it. The motor shafts are 3/8" on one end and 1/4" on the other, the ballscrews are 5/16" and the only handwheel I had is one from my Sherline systems, which only fits 1/4".

I started out about noon to make a quickie adapter, so that I could put the 1/4" drive Sherline handwheel on the 5/16" ballscrew, but realized that was overly complicated. I have the original Grizzly handwheels, and they have a 0.393" ID, so I all I needed was a little steel sleeve adapter to make it fit the 5/16" shaft. The Grizzly has a Woodruff key, which I replaced with a setscrew.

With this on the left end of my X-axis, I can wobble the handwheel quite noticeably before any motion starts. There's my .012 backlash, I bet. Of course, I need to take off the motor and all on the right side, but with this in place, I'll be able to feel the backlash come out.
Adapter.JPG
Unfortunately neither the Y or Z axes have a place to mount this on the far end while I mess with the driven end.
 
As a side note, I may as well ask here. Is anyone using LinuxCNC on a system like this? Anything using Chai's C7 ballscrews?

I'm getting the most bizarre results with LinuxCNC's StepConfig program to determine accuracy and would like to compare the numbers for leadscrew pitch that work in the StepConfig window. The program asks for turns per inch. I set my calipers to 1" and counted the number of threads as 5. That was close enough to allow me to tweak that value to get the table travel to be correct.

When I first started moving the table, it went much, much farther. I had to tweak that number down to 1.27 TPI from 5! The thing is, when you turn the handwheel, you get almost 5 turns.

These numbers work out if you convert the 5mm pitch (RM1605 or 2005 ballscrews), one turn is 0.197", which is what I get. That turns into 5.08 TPI. A far cry from 1.27 TPI.

Thought I'd ask here before I went to the LinuxCNC forums.

Backlash.JPG
 
Did you set the steps per turn?
 
Back
Top