Hello everyone.

My machine is a 60mmx40mm desktop CNC, upgraded with an ethernet board and closed loop steppers and stepper drivers.

Beforehand, I want to state that I have done the 300mmx400mmx500mm XY plane calibration method, and measuring with a digital caliper numbers were within less than 0.1mm in both axis, including the hypotenuse. That means steps per mm are well adjusted over a somewhat large area.

I recently found out that my desktop CNC is doing overcuts on the Y axis of about 0.2mm. I was measuring some jobs that included 90mmx90mm square shapes, and on the X axis measurements were pretty much spot on, with about a 0.05mm difference measured at different spots.

Y axis though gave me on these 90mmx90mm squares an overcut of about 0.2mm, measuring 89.8mm on different spots, pretty consistent on all of them.

I thought it was due to my machine needing calibration once again, but to be honest I haven't pushed it cutting soft metals or anything hard since a very long time ago after last calibration. Still, I went ahead and did the 300mmx400mmx500mm XY plane calibration method once again, finding there was no issue, which meant steps per mm were fine.

To discard more issues, I went ahead and cut a 150mmx150mm square on some MDF wood, and to my surprise X axis was spot on, and Y axis still had that 0.2mm difference, measuring very consistently about 149.8mm on different spots.

I don't see any tool offsets on Mach3 and I have used Gcode from different software getting the same results, which also means it is not some offset being placed when generating Gcode, which would be visible anyways in the software settings and output files.

Also if it were some sort of skew, sides would be angled but consistent, and they are not angled, if it still were a steps per mm issue the error would increase as the piece gets larger, but it is a consistent 0.2mm overcut.

In other words, work is getting shrunk by 0.2mm on the Y axis, and yes, stock is well fixed to the machine bed.

Any ideas why this might be happening or has anyone experienced something similar?

Have you eliminated backlash?

It could be something loose or backlash or calibration or the head is moving back and forth on the rail.

Small aluminium frame machines are not rigid so they do flex, I would start by checking for movement in the Z axis.

Have you eliminated backlash?

Well, this machine came with anti backlash couplers, the ones that attach from the motor shaft to the leadscrew, and the also have these anti backlash mechanisms that go in the recirculating ball screw mechanism, the ones that you tighten with an small Allen screw, I don't know how these are called exactly, so I don't know if backlash it's quite the cause, maybe it is, but wouldn't know how to measure it in a better way, doing travel tests give me a very spot on measurement, probably out of luck, but wouldn't be sure either...

It could be something loose or backlash or calibration or the head is moving back and forth on the rail.
Small aluminium frame machines are not rigid so they do flex, I would start by checking for movement in the Z axis.

I have checked if the spindle flexes by hand and a Mitutoyo dial indicator, and that is by applying A LOT of  force. If that were the case, the piece would throw random measurements bigger than what they are supposed to be, not a constant 0.2mm less every single time, and yes, it is a small machine but the frame is not the super cheapo one. I have disassembled it before and changed all Allen screws with better ones tightening all of them firmly back when doing the stepper motor upgrades and tramming adjustments, plus, I have made the tests I mentioned on PCBs and soft wood with a maximum of 0.1mm depth of cut so flexing would be discarded.

If it were backlash, is Mach3's anti backlash system any good? I have digged a bit on the net and they talk about changing acceleration settings and some sort of sudden stop, but wouldn't know how to configure them correctly to make proper tests without ruining default settings that have worked fine until now, if that were really some sort of starting point to make tests...

I have anto backlash set up on my machine correcting something less than o.1mm IIRC.   But first you have to measure it with a dti and careful jogging.

Try cutting squares from 10mm to 100mm and see if the error changes as the size increases.

I have anto backlash set up on my machine correcting something less than o.1mm IIRC.   But first you have to measure it with a dti and careful jogging.

Well, I did backlash tests on my Y axis using a Mitutoyo dial indicator, and backlash was only between 0.02mm to 0.03mm. So it is definitely not a backlash problem, since the measurements I mentioned before go down to 0.2mm less (10 times larger of a cut).

Try cutting squares from 10mm to 100mm and see if the error changes as the size increases.

I have done so before as well as mentioned previously, cutting error doesn't increase as the piece gets larger, it is a pretty much constant 0.2mm overcut.

What could be next?

After looking around at what may be the cause, it may be definitely related to backlash, so I went ahead and looked at the insides of the back panel for the X axis and the parts below the Y axis.

For the X axis, there is a HUUUUGE one-piece slide bearing block. The center hole for the leadscrew and four more for rod supports.

This whole pillow block has tiny grub screws to adjust backlash on each rod, two grub screws for each rod, meaning there are two ballscrew bearings for each rod. There center whole has no apparent way to tighten it, just a grub screw for what seems to be a white cap to prevent dirt going in.

The bottom part for the Y axis has four pillow blocks, two on each side for the support rods. Also, there is a block for the middle part where the leadscrew goes, with no apparent way to tighten it as well.

To see how it would affect the X axis, I stupidly toyed with the adjustable grub screws, and man did I mess up. I was getting perfect measurements for the X axis and now it is having the same problem as the Y axis, so yeah, the solution seems to be there, but no idea on how to properly adjust them in order to get a great result.

I have looked all over, but all I see are ways to adjust those blocks before mounting the entire chassis, and yes, after toying with them I see how backlash is affected, but it either gets worse or I manage to get it back to the same amount of backlash. If too loose, I don't get good results, if too tight, I would've thought there would be no backlash, but there is still is, and the machine sounds way too rough...

So, does anyone know a proper way to adjust theses pillow blocks without having to dismount the entire machine?

Any chance of some photos please?

Any chance of some photos please?

Sure, these are the pictures of the X axis block I mentioned, the what seems to be a white cap, and the bottom of the machine which is the Y axis.