I have always had an iffy z axis. So as a test, I just cut a square 0.1" deep. After the cut, I measured the depth of cut and the starting point was indeed 0.1. But the ending point was a depth of 0.15 and the surface of the material was no longer 0. Then I reran the gcode so it was actually cutting air. This time it appeared to cut at the proper depth and zero was still at the surface of the material. I have thought my table was not flat was why my cutouts went through on one end but not the other. However, I zeroed z at the surface, then raised z and jogged x & y about 12 inches away and lowered z back to zero and it was within 3 thou. So the table seems flat.
I know I am not bottoming out, the backlash is set so I can still turn the z by hand (with effort). The two pulleys are tight. I read I should tension the z axis drive belt, but I can’t figure that one out.
@Bruce Welcome to the group, Bruce. You will find this place an excellent resource, I am sure.
You say that you can’t figure out how to adjust the Z gantry drive belt. Take a look here:
Part way down, you will see a reference to belt tension.
With reference to your problem, if I understood your post, everything is fine when you run in air. The problem arises when you are cutting. I would agree with @JHahn that the bit may be slipping. Are you using a router or a spindle ? Does the problem arise with all bits? If it is present when using 1/8" shank bits, are you using a 1/4" to 1/8" adapter or the 1/8" collet.
The one I was currently using is a 1/4 and I had cleaned out all of the sawdust before inserting the bit. I reviewed and reinstalled the z axis pully belt ( it did seem as though it may have had a little slop). I accidently broke the bit I was using, so I will be now using a new one. I measured the shaft and it came out to 0.2495, so it was within 5/10,000. so if the problem now goes away i will be happy, but not know if it the bit slipping or the lose belt on the pully.
As the original issue has been addressed, I’m closing this topic. If anyone else has a similar concern, feel free to start a new topic, which will ensure that it is addressed promptly.