Question concerning steppers and belt drive

[kae]

I am finishing a CNC conversion of a square column (Precision Matthews) mill. I am using 20mm (5mm pitch) Rex ballscrews and 600 oz/in Rex steppers from homeshopcnc. I built a 2:1 ratio (20 teeth on motor, 40 teeth on screw) timing belt and pulley drive to double the effective torque as well as increase resolution on each axis. I am using 203v Gecko drives so I am microstepping at 10x. I calculated the steps per unit (inches) at 20408 steps per inch. Using Mach3, I set the steps per unit, chose some reasonable acceleration and velocity and did some test runs at one inch increments. I have linear glass scales installed so that the machine can be run in manual mode for simple operations. Using the linear scales as standards, I discovered that there was a linear error in the calibration. I was gaining about .0035 per inch of travel. I read in the Mach3 manual that you can tweak the steps per unit to bring the system more into calibration. I reduced the steps per unit from 20408 to 20340 and now I hit 1.000 pretty close but I am still experiencing a linear error of .004 in 8.000 inches. I am attributing this to the natural error in a grade 7 ballscrew even though Rex indicates that these are good for .0018 per foot.

This next problem is a little different. If I do G0 rapids out to some distance then use the Mach3 "Return to Zero" it will return dead on only if I return to zero from the same direction that I set the DRO zero. However, if I return to zero from the opposite direction there is a small error. This is true for both X and Y axis. My first thought is that there is some slight "backlash" in one or both of the following: timing belt/pulley assembly (the belt shifting slightly between teeth when it changes direction) or some backlash in the ballscrew/ballnut even thought these are suppose to be anti-backlash. I feel pretty good about the preloaded AC bearings I used on the ballscrews.

I would appreciate any comments on either subject based on your own personal experience.

[TOTALLYRC]

I seem to find that the calculated value is close but always requires tweaking to really get it where you want it.

You can put a decimal point into the steps per window to really dial it in.
On every machine that I have set up I have had to put the steps per in with a decimal point.

I would use a dial indicator to make sure the "Backlash" isn't in your glass scales or their mounts.

Mike

[Pandinus]

For the steps per inch, go to the settings page in Mach3, just above the reset button is Axis Calibration. Go to zero on your DRO first, then click the button. It asks axis? and how far to move. Enter the max your DRO can measure on a full move, then MAch asks how far it moved... Enter Whats on the DRO and accept the value it calculates for you.

Try moving a few thou, then back. If the pulley doesn't rotate more than a few degrees, I'd assume the belt didn't shift, if the tension in sufficient. If the backlash is the same its somewhere else, but if its gone your probably correct with the belt shift.
How much drag is there in your ways??? Whats the diameter of the pulleys?? How difficult is it to turn the screws??? If its a fine pitch = smaller diameter and a narrow belt then the belt may be distorting a little during movement. Just a guess tho. Check if one side of the belt is looser after a move in each direction. If it is and it changes, tighten the belt a little or loosen your ways... you don't want to introduce error in your axis via slop either.

[kawazuki]

The one thing I picked up in the OP was 'anti' backlash screws! Not knowing your set up and assuming you are not using double ballnuts - some backlash is enevitable. 'Anti' only means resists, it doesn't mean eliminates, for that, it requires a Zero backlash description IMHO
With a 5mm screw, I calc your steps per as 20320 anyway so you were off to start with?
25.4 / 5 = 5.08 x 200 x 10 x 2 = 20320 steps. If you really are at 0.0005 / inch, i'd use backlash comp in mach3...............but then i'm cheap! LOL

[kae]

Thanks guys. I'm not surprised that the systems require some tweaking. I'm a little embarrassed that my math was off. I appreciate all the suggestions and comments. It gives me some things to try. I will try to post some follow-ups. It may help someone else along the way.

[kae]

Your comment on the anti-backlash nut is well taken. It seems to be a managable number and I will try the compensation. I prefer frugal

Originally Posted by kawazuki The one thing I picked up in the OP was 'anti' backlash screws! Not knowing your set up and assuming you are not using double ballnuts - some backlash is enevitable. 'Anti' only means resists, it doesn't mean eliminates, for that, it requires a Zero backlash description IMHO With a 5mm screw, I calc your steps per as 20320 anyway so you were off to start with? 25.4 / 5 = 5.08 x 200 x 10 x 2 = 20320 steps. If you really are at 0.0005 / inch, i'd use backlash comp in mach3...............but then i'm cheap! LOL

[kae]

Mike,

I checked the scales against a precision standard and they are dead on. I have been tweaking the calculated value and I keep getting closer. Had to stop for the evening. I do have some backlash. I now suspect the screws but it could be a combination of things. Thanks for the input. This community usually leads me in the right direction.

Regards, Kent

Originally Posted by TOTALLYRC I seem to find that the calculated value is close but always requires tweaking to really get it where you want it. You can put a decimal point into the steps per window to really dial it in. On every machine that I have set up I have had to put the steps per in with a decimal point. I would use a dial indicator to make sure the "Backlash" isn't in your glass scales or their mounts. Mike