Weird size deformations

I’ve been playing around with trying to calibrate my mpcnc. Overall, 50mm squared cut out of mdf comes out 50.2x50.2 with angle not being square by around 0.2mm over 50mm, which is more or less fine for my needs. After getting these results consistently, I tried cutting out a hexagonal box. In theory, any diagonal should’ve been 86.6mm, in practice I got extremely weird results.

Every diagonal is different by a significant amount. And all are far from the desired size. My previous tests were out of mdf and this is wood, so grain orientation might be an issue here. I’m using the same feeds and speeds both on mdf and wood: 12mm/sec, 1mm doc.

Start with a pen to see if the inaccuracies are the machine or how hard you are pushing the bit through the material.

If that checks out make sure you are using the correct sized finishing pass.

1 Like

Being out of square (a bit) is already enough to produce different corner-to-corner diagonal lengths on a square. So the different diagonal lengths between the faces of the hexagon is not too weird. It’s just a question of squareness and dimensional accuracy, which it sounds like you already know how to measure.

Ok, I need help here… =(
I’ve decided to step back a bit and try a few hexes. A hex of 80mm between sides. 10mm/sec, 1mm stepdown, finishing pass of 0.3mm + spring pass. Had 80.0 on axis, 80.2 one diagonal, 80.4 another diagonal.
I though that my mpcnc was not squared good enough - I remember when assembling it I just used regular ruler without too much trying. So I decided to disassemble everything, square it and try again. This time I did the best I could - I was measuring loop around tubes instead of distance between bases, since full loop measures twice - I get twice the accuracy, so I’m pretty sure that both sides and diagonals are within 0.5mm.

I’me using mechanical stops to square the gantry during turning on.

I’ve measured D1 & D2 with digital calipers and they are within 0.1mm…

After that I’ve tried running the same gcode - - and to my surprise I got the same sizes. I.e. 80.0, 80.2, 80.4 - within 0.05 according to calipers.

I’m stuck here… why are sizes so different?

Are you using a full depth finishing pass?

Hard stops are not even close to as accurate as dual endstops.

3d printed pulleys?

Under 0.4mm is pretty good. And that type of spindle is known to be pretty sketchy. What I am getting at is that last bit of accuracy is very hard to track down. You really need to look at everything.

I’m doing full depth finishing pass.

I understand that hard-stops are kinda meh, but I don’t have a controller board capable of driving all 5 motors atm. Planning to order pcbs soonish…
Yeah, 3d-printed pulleys, factory-made ones should arrive in a week or so, will test them as well.
Spindle doesn’t have visible play or runout, at least visually.

I would understand if I was getting 0.2-0.4 randomly. But this result is repeatable (actually, I will try to cut another one just to make sure I still get the same) enough to make me suspect that there is something I’m missing…

What does “during turning on” mean to you? Mechanical end stops like that would require you to hold the gantry against it securely on each end while engaging the stepper motors. Not an easy task. Just turning on the MPCNC isn’t enough.

To me most of the time pull to end stops start machine move towards stops till both sides skip step then check for square and then set 0 0 0. Hopefully in the next few days I will have end stop working

yes, grbl have hold axes enabled on turn on, and I am pulling both axes until they are firm against hardstops… my mpcnc is on a smallish side (300mm squared work area size), so I can see and control that 4 are there.

Belt tension equal, zip ties equally constrained?

Belt tension equal, zip ties equally constrained?

well, belt tension feels equal, zip-ties have those loops around them to constrain, but I wasn’t over-zealous in trying to make sure that belts are absolutely equal-length, so loops of the zip-ties might differ a bit.
Will get new pulleys somewhere mid-next-week, will try again.