[webgeek]

As I've documented in my previous thread, my X2 accuracy is quite poor (on the order of a 1/10" off at times). This is likely my fault in how I put it together. To be honest, I've not touched the machine in months as this process scares me for lack of a better term. Every time I've touched the machine, I've ended up making it worse. In the last few months, I spent my time getting good at CAD and such and now I'm up against a wall. I need to start making chips this week or I won't be able to finish my robotics project in time for the competition I'm attending in a few months.

So, back to square one. I'm trying to fix the accuracy problem. I've torn the machine all down to just the Y-axis components, re-tightened them down, and then lubricated it all up. I've attached a dial test indicator to the chuck and then jogged off and on to see if the same value is consistently repeatable. As best I can tell, it is. Here is a video showing what I've done:

[ame="http://www.youtube.com/watch?v=62mPXXWuo_I"]YouTube- Y Axis Test #1[/ame]

So my question is: what should I check next? I'm trying to put each axis to bed as I go. The Y axis didn't appear to be much of the problem the first time around but I figure it's best to test it and be sure. Also, what should I be putting locktite on? Someone suggested locktite before but I've not used it on anything yet for fear of putting it on the wrong part. Please also remember I'm a total neophyte so try and keep it simple Thanks much for any help!

-Mike

[hoss2006]

I've always suggested folks use Loctite blue 242 threadlocker on the mounting screws for my conversion parts.
It's not permanent, just keeps the screws from vibrating loose.
Can't help you calibrate the mill with EMC2, I'm a Mach man.
This older calibration video I made could show you how to test your system to
make sure that the software (Mach 3 in my case) and the machine are moving the same distance.
I would imagine you can adjust your steps/per in EMC2 as well to compensate if necessary.
Hoss

[ame="http://www.youtube.com/watch?v=OTHdgk12EcI"]YouTube- Mach 3 Steps Calibration[/ame]

[webgeek]

Thanks for the quick reply! Sorry for not being more clear. I've got EMC set up properly, I'm talking about the mechanical side of things. Prior to this latest tear down, if I told it to cut a circle, it had obviously flat spots on both axis. Additionally, the ball screws were a bit noisy. I'm just trying to make sure I've got the mechanicals accurate each step of the way before I finish putting it together.

-Mike

[cyclestart]

Beware of the step and direction timings provided by emc for the G540. I was getting some wild results (posted in the emc subforum today) before realizing what was wrong. Your problems might be mechanical but don't rule out some misconfiguration.

[H500]

If you're .1" off, then I doubt that it's mechanical. It sounds like your controller is missing steps somehow. Try the jogging the axis back and forth for a minute and then see if it returns to the same spot.

[webgeek]

If you're .1" off, then I doubt that it's mechanical. It sounds like your controller is missing steps somehow. Try the jogging the axis back and forth for a minute and then see if it returns to the same spot.

That .1" was measured when a circle was pocketed out. The top/bottom and left/right sides were badly flattened. Lost steps seems like they would show up very differently than a consistently flat edge? Either way though, I went ahead and did the test you suggested and it came out perfectly, with no loss of precision that I could see.

[SCzEngrgGroup]

That .1" was measured when a circle was pocketed out. The top/bottom and left/right sides were badly flattened. Lost steps seems like they would show up very differently than a consistently flat edge? Either way though, I went ahead and did the test you suggested and it came out perfectly, with no loss of precision that I could see.

Flat sides when making a circular pocket is a clear indication of backlash. 0.1" is a HUGE amount of backlash, which should be very easy to find. Something is VERY loose - shaft couplings, motor mounts, bearing mounts, leadscrew nut/ballnut mounts, etc. It just can't be hard to find such an enormous amount of slop by simply pulling and pushing on various pieces of the system until you find something that moves.

Regards,
Ray L.

[webgeek]

Flat sides when making a circular pocket is a clear indication of backlash. 0.1" is a HUGE amount of backlash, which should be very easy to find. Something is VERY loose - shaft couplings, motor mounts, bearing mounts, leadscrew nut/ballnut mounts, etc. It just can't be hard to find such an enormous amount of slop by simply pulling and pushing on various pieces of the system until you find something that moves.

Yup, that was my understanding and why I've torn it down to the Y-axis alone. As of right now, I can't find any backlash in the axis. I've gone over it with a fine toothed comb and tested it multiple times and the axis seems consistently good now. The X-axis is lubricated up and I'll be putting it on in an hour or two, to test it. Thanks!

-Mike

[webgeek]

So after tearing the danged thing apart and putting it back together again, it's still pretty mediocre. There are clearly visible flat spots in circular pockets. This is driving me absolutely crazy. I'm not an accuracy nut but this is pretty pathetic.

I'm sure I've screwed something up again but I have no idea what. If I tighten things down anymore I actually prevent the axis from even turning. Everything is lubricated, everything is tight and cinched down, etc. Here is a picture showing two circular pockets cut in wood to see the issue.

The picture actually looks a touch better than it does in real life. Measured top to bottom, the outer pocket is .987" and from left to right it is .998". Measured diagonally from upper right to lower left, it's 1.014". Obviously it's cut in wood, so I expect it to be a little inaccurate due to the material but not this much. It's basically .014" off. I could live with something that at least LOOKED round - so if I could get it down to .07 or so I'd be satisfied, if not pleased. As it stands, it's off enough that my screw holes in parts are likely not going to even line up.

Every test I've done makes it appear to not be missing steps. It's not mechanically missing steps either as that's quite audible when it occurs. This specific shape distortion is quite repeatable so I believe it's a mechanical problem. The ONLY thing I didn't re-cinch down is the Lovejoy couplers on either of the X/Y motors or ballscrews. I didn't do it mainly because they are a horrid pain in the A** to get mated together. They were very tight previously as I had screwed em down (and added flats to the ballscrews) during the last tear down.

I'm out of ideas here. Any suggestions? I appreciate the help as always. At this point I'm rapidly starting to regret the decision to CNC this machine in the first place. I have no interest in fussing with the machine constantly, I just want to cut 2d parts I design in CAD. To be honest, I might have a converted machine for sale here pretty quick at this rate as this is proving to be a waste of money and is rapidly giving me high blood pressure.

Thanks again for any advice!

-Mike

[eartaker]

2 things come to mind when I read this.... Backlash and steps per inch. you may have a small about of slop in your shaft connectors or on the ball nut.. If not then it is all software. I use MACH and I'm sure in your software there is a way to calculate the steps per inch rather than doing your own math. Sure you can take the thread pitch and the steps it takes for your motor to make 1 revolution and figure it out that way BUT.... the software is much more accurate. My machine had the same issue and it was all software.

[photomankc]

I'm a newb here but I was thinking you might setup a 1" Dial indicator so it's at it's halfway point on the axis. Then jog out .400" and see if it really moved .400, then jog back by thousandths and see how much backlash there is before that dial starts moving the opposite direction. Everything I have seen so far seems to say that flat circles are a classic sign of unaccounted for backlash.

I have a test rig I'm using to develop my own driver. I just setup your test. I put my indicator on the axis after backing up into position and started at 21. I jogged forward on the test board by 0.010" and I moved to 25. So 0.006" of backlash. I then moved 0.010" back and I landed right back on 21. This repeats every time. So coming back to the same point is not proving the backlash is gone. Telling it to move forward by 0.010" after moving backward and having it move exactly 0.010" forward would tell you if EMC is really taking up the backlash when reversing direction.

Edit to add: My microcontroller board currently has no backlash compensation... just to be sure thats clear so it can't have any role in making sure I land back on the correct position when jogging out and back. Apparently that will happen even without compensation.

[hoss2006]

Yes you have backlash and most likely it's coming from the lovejoys,
unlikely you can do much other than replace them with zero backlash couplings.
this video is again using machs backlash comp, but the procedure for verifying it on the mill will be the same.
You want to get everybit out that you can mechanically, comp should be just for the last few thou that won't go away.
Hoss

[ame="http://www.youtube.com/watch?v=buXHZ_McbOY"]YouTube- Verifying Mach 3 Backlash Comp[/ame]

[LongRat]

Using your dial indicator to read the table motion, you sould be able to ascertain the backlash with/without the Lovejoys. Jogging back and forth and reading the lost motion on the indicator 1st, then disassemble the Lovejoy and do the same thing, turning the screw by hand. The amount of backlash you have is so big it will be obvious that there's lost motion even turning the screw by hand. You will be able to feel it. You can also lock the table gibs and turn the screw by hand to feel the backlash. If you can't feel a large amount (1/10 of a revolution of the screw for your 0.5mm lash if you have 5mm pitch screws, for example) then the problem is definitely in the Lovejoys.

For the record, clamp type are much better than set screw.

[webgeek]

Well, I was wrong. The lovejoys are not the problem. The X-axis has about .003 of backlash currently and it appears I've been able to get EMC to compensate for that. I didn't bother trying to get to the bottom of it as it wasn't that much.

The y-axis is a totally different matter. It has a whopping 0.015" of backlash. The backlash all appears in the ballnut's connection to the ballscrew as best I can tell. The ballscrew has a good bit of turn before the ballnut engages. I determined this by rotating the ballscrew by hand and seeing how much movement it had prior to engaging. The lovejoys and the rest have no noticable backlash, it's all in the ballnut/ballscrew. This also explains why tearing the machine apart and putting it back together several times has made no difference.

Considering I paid a bit extra for pre-loaded ballnuts, this seems REALLY excessive. I've never taken them apart and they are tightened down and lubricated as well.

Should I be contacting CNCFusion at this point? I don't know how to fix the ballscrews and I'm nervous about making it worse. Thanks!

-Mike

[LongRat]

That's not what I would call an acceptable amount. If it is in the nut-screw connection, I would send it back.
I am assuming that you have put an indicator in-line with the end of the screw and determined that it isn't moving axially? And the nut mount to saddle connection is tight.

[BobWarfield]

With 0.015, I would think you could jiggle the ballnut on the ballscrew there's so much.

They use oversized balls rather than 2 ballnuts?

If so, looks like the wrong sized balls got loaded.

Cheers,

BW