Howdy Going through the late stages of converting my RF45 clone to CNC. I'm using DMM Technologies servo drives and 400W motors, 3:1 belt driven, with the typical 5mm lead ballscrews.

I have noticed while tramming and measuring backlash, etc, that while jogging the machine in 0.001" step increments, the axis' don't always move a full thou.. The servos have 500ppr encoders, with quadrature, which turns out to be 30480 steps per inch. Even ignoring the quadrature, it should give me step accuracy of 0.0001" or thereabouts, but I'm not seeing that. It definitely averages itself out, if I move 0.020", it'll end up fairly close, but some steps will move it 0.0015, some are right on at 0.001 and some 0.0005.

Any ideas? It could just be horrendous ballscrew accuracy, but I thought I'd check if anyone has seen a similar issue. Any thoughts appreciated

Thanks!

edit: figured I should mention that I'm also using an ethernet SmoothStepper!

Put a piece of tape over the 4th digit right of the decimal on the dros?    

In all seriousness if your steps per are 30480 you should be able to move with that reseloution or pretty close.  Its best if your steps per are evenly devisible by your desired reseloution. 

 

if I move 0.020", it'll end up fairly close, but some steps will move it 0.0015, some are right on at 0.001 and some 0.0005.

What is reading the 0.020, 0.0015, 0.001 and 0.0005?  Where are these numbers coming form?  A dial indicator?  Mach DROs?

Brett

I knew I had forgotten some little detail this was measured with a test indicator.. To be fair, if half a thou was all the error I got with the machine, I'd be doing great! Just trying to sort out whatever I can..

I have also noticed that backlash compensation works a bit randomly - lets say I move in X+ until my indicator reads 0. I had 0.003" backlash, set comp to that, etc. if I step in the X- direction, the indicator will sometimes not move on the first step, then move 0.002" on the second step.  This is measured on the indicator, but not reflected in the DROs..

Backlash could be why your seeing random moves and will likely cause you a number of headaches if your trying to hold tolerances less than the backlash the machine has.  BLC is a work around for inferior mechanical components.  You will probably be much happier if you concentrate on getting rid of the backlash.  Also, if your ballscrews are metric, I would set the native units to metric and set everything up for that.  You can swap units back and forth easy enough between metric and imperial code execution.  That isn't mandatory but my preference.

Brett

 I have to disagree regarding the native units I am afraid, I would set up in the units you will predominantly use, shouldnt be a problem for Mach and might even be a benefit.

I agree regarding the backlash though, turn it off, at least while you are testing, and see if things are better. The other thing it could be is your axes friction/stiction and the tuning of the servo drives, if the tuning is loose and a wee bit of stiction on the ways then the drive may be quite happy with the position it has seen from the encoder as it is well within its error limit. This sounds like it may be the case as you say on a longer move it is accurate.

Hood
Hood

Thanks for the replies, guys. Yup, backlash is definitely evil.. Preloaded double ballnuts are on the to-do list, but it will likely be a few months before I get around to that. In the meantime, I'll play around with disabling backlash completely to see how that'll work. I don't believe the machine in it's current incarnation will be accurate to much more than a couple of thou anyway, so it's possible that disabling backlash comp won't introduce that much error, but will provide some other benefits.

Hood, that is a great point about being within the error limit of the drives. I have tightened up the gibs significantly to get around a generally poorly manufactured machine. The following error on the drives is 127 steps, which means at 30480 steps per inch, it's a little over 0.004". I'm sure the drive tries to stay as close as possible to the correct position, but like you say, it's very possible that it gives up if it's close enough.

"I have tightened up the gibs significantly to get around a generally poorly manufactured machine." - That would be my first guess as to your problem.  What you describe are classic signs of "stiction", which means the axis is not moving as freely as it needs to for precise moves.  You have to tighten the gibs WHILE watching what backlash is doing, which means starting out loose, and measuring backlash.  There will be a lot.  Now tighten the gib a LITTLE, and measure backlash again.  Keep going, and at some point you will see backlash start to decrease.  Keep going, and at some point, you will tighten the gibs a little more, and backlash will *increase*.  At this point, friction is preventing the axis from moving freely, and tightening the gibs further will cause backlash to increase rapidly.

Regards,
Ray L.

Thanks Ray, that's a great explanation. I'll give that a try as soon as I can!

My gib strips are far from perfect, I had to make them very tight to get rid of side to side play in the table/saddle. Long story, but it's a pretty old RF-40 table/saddle/base. I have a new RF-45 table on the way to go with my spare RF-45 saddle and base, which I'll take the time to match as well as I can. I'll have to make a new tapered gib and most likely scrape it to fit. Depending on how well that goes, I might scrape the saddle and table too, but I'm really not looking forward to doing that