Bizarre Backlash Problem?

I'm having a problem that has me completely stumped. I have a Taiwanese Bridgeport clone knee mill that I've recently CNC'd. I used double-nut ballscrews, big servo motors, and timing belt reducers. The problem is a backlash anomoly that I can't explain. I decided to check the backlash a few days ago. I put a test indicator on each axis, and used Mach3 to move first one direction, then the other, in 0.001" increments. I found what appeared to be 0.0028" backlash in Y, and 0.0060" in X. i.e. - any time I reversed direction on an axis, Mach3 would move a few thou (and the motors *were* moving) before the indicator showed any movement. So, I dialed compensation into Mach3, and was then able to move consistently and repeatably over short distances on both axes. However, when I moved a large distance (say 10"), the move would come up long by roughly the amount of the apparent backlash.
I then tried cutting a circular pocket, with backlash compensation enabled, and it showed the signature "divets" at 12, 3, 6, and 9 o'clock. I turned off the backlash compensation, and the pocket came out pretty much perfect all the way around.
So..... How come I measure backlash when I check with two different indicators, yet in actual use, there appears to be zero backlash? I'm sooooo confused!

Regards,
Ray L.

Hi,

You could have a case of sticktion combined with a spring action. If you can get
an indicator on the end of the ballscrew and see if it is moving. You may find that
moving small amounts the screw/thrust bearings are giving way and spring back
into proper position when the slide overcomes the sticktion of the way surface.

It could also be occuring between the double nuts. If you can mount an indicator on
the table and place the tip into the thread of the screw, you should see no motion
when you move the axis.

Hope this helps

What did you use for thrust bearings on the screws? Deep groove ball bearing = bad, angular contact = good if properly mounted and loaded.

Are the ballscrews precision ground or rolled thread? The difference this will make is in the true travel distance of the raceway over a long distance, as well as the degree to which you can preload the double nuts is limited by the screw accuracy.

I've put an indicator on the leadscrew end, and it's not moving measurably. Also, I check the ballscrews on the bench before putting them into the machine, and saw zero backlash. The bearings are matched pair angular contact. The screws are Nook XPR rolled screws, 0.001"/ft. lead error - only 2X the error of ground screws. This problem is evident over the full length of the screws, not just at one position.

Regards,
Ray L.

One other thing to check is the tightness of the slideway gibs. On a knee style mill, looseness in the Y gib can affect the X axis and vice versa. You can often tell by placing and indicator near the extreme end of travel and pushing or pulling on the table in the appropriate direction and see how much slop you'll get. You'll probably have to put up with a little bit, but the gibs should be as snug as will permit full table travel in each axis.

My knee mill will act similarly when the timing belts are getting worn and/or loose.
Maybe double check the belt tension and make sure everything is tight on the pulley.
I don’t know what type of servo drive system you have but have you checked for positioning errors in the drives themselves?