Random loss of accuracy?

I’ve run into an odd situation and I’m curious what’s causing it. I use dial indicators to calibrate my machine. A few weeks ago I had my machine dialed in within 0.000394" on all axis. I milled a piece of aluminum and ended up with a .02" undersized hole, which is crazy innacurate. I calibrated again and sure enough I measured a .0197" deviation. Could this just be belts stretching? I’m using 9mm 3GT3 belts. This is starting to make me concerned the machine can’t hold it’s calibration.

Did you measure across the same distance in the same spot on the machine?

The first time I was measuring 250mm travel at two locations on the X, and two locations on the Y. I thought I was pretty thorough. It’s the same way I calibrate my other machines. When done I measure a 700mm travel, and three travels of random distances, home the machine, and then do the 250mm test one more time. Which, I feel is a rather elaborate and unnecessary for any properly calibrated machine. Either way, losing .02" of travel is just crazy. The most concerning part is my Z axis losing .02" of accuracy, since it’s a linear bearing screw drive that has repeatable positional accuracy of 0.0002" with very minimal backlash. That should never happen, or at least to that kind of degree. It made me wonder if it was something software based. I did recalibrate last night, and I’m going to run some more tests this evening.

X, Y, & Z all lost .02"?
Is the Z loss measured through actual movement or in measuring the part?
My reason for the questions is that I’m wondering if the aluminum could introduce enough deflection and/or flex in a machine that isn’t very stiff.

That is exactly what I thought as well. No, this occurred in movement, measuring jogs. I checked for deflection by measuring the tops and bottoms of pockets, and it was within negligible amounts, no perceived deflection (or very minimal). The only thing that makes sense to me is the belts loosened, because everything seemed to happen uniformly. That doesn’t explain the Z axis, though. I was using a digital caliper to measure the 250mm travel, and it is possible my caliper had some inaccuracy (but .02" seems unfathomable - it would be causing screw ups all over my shop). This last time I pulled out my Mitutoyo digital dial caliper to remove any doubt. My Belts were all tensioned to 3.5lbs about two weeks ago just before I did my calibration. Maybe temperature changes could have explained it - I’m about 10 degrees cooler, but I really don’t think a Kevlar belt is going to have any discernable change in tension on such a minor temperature variation.

Now that I think about it, I did make some rapid and speed adjustments after those calibrations. I wonder if that somehow could have affected the accuracy? Loss of steps or something related? None of my speeds are very aggressive, so I can’t imagine it would, but maybe. I’m just trying to think if there is some setting that might have some sort of an influence over all axis somehow.

If my brain is still working at this hour, At 40 steps/mm, you’d need to lose 20 steps for a 0.02” loss.

250mm jog is 10000 steps so 20 / 10000 = 0.2% of steps.

That’s a pretty tiny number of steps that you’d probably only notice if you’re measuring with a caliper.

Just a small slip or stall due to acceleration or velocity could cause it.