I tried making a quick manual code to test for backlash. It runs in a square, but halfway along each side it stops, turns off the laser, runs to the opposite side of the square, comes back, then keeps burning the line. That way it will make all 4 +/- x/y movement changes (so the thinking goes).
Backlast Test 1.nc (802 Bytes)
The thing is - it came out suuuper clean. I can’t see any missed movements, so do I not have backlash, proof positive, or is perhaps this test not doing it’s job?
Edit from mobile: pic upload
Units for the ruler here are decimal inches (smallest demarcations are 0.01")
I’m stewing on your test file and I don’t think that file actually tests backlash. Maybe I’m missing something so could you explain which specific lines or movements are testing for backlash? Remember backlash only occurs when the axis direction changes. If you approach from the same direction backlash is removed.
The sides of the square are always burned with the same direction leading up to them. On the first side, for example, the laser burned in Y+, then again in Y+. On side 2 the laser came in heading in Y+ and heads off in X+. After the jog the laser is burning having approached from Y+ and is heading in X+, the exact same.
That move sequence is what you would generate if you’re intentionally removing backlash by preloading from the same direction always.
The only possible locations for backlash in this I think would be at the origin, and if the backlash is smaller than the kerf it would be hard to determine.
I think a better test for Y backlash would be something like this:
If there’s any backlash in Y there will be a visible discontinuity between the end points of 1 and 2 because they approach from Y+ in 1 and Y- in 2.
You could then add in a cross for testing X but careful cause I can’t figure out how to draw that without making a swastika 
I did a circle because a single circle will measure X and Y backlash on the boundaries between the 4 quadrants, but it works best with multiple tests at different compensation values where you can pick the middle best - hard to pick out by eye if a circle is slightly off of perfect by a few hundredths of a mm. Unless your backlash is so huge you end up with something like this, which yours is not:
Thanks man, I didn’t give any thought about which way I was having it jog. I flipped all those sections so that it runs (for example) x+ into the corner, then a jog of x+ leading into x-, so that the second half of each face of the square is approached from outside, while the first half is approached from inside.
And yeah; found it!
And here’s an annotated version:
Plus some zoomed shots of the backlash
X axis:
And Y axis:
Which also totally shows some slight wobble, which I’ve noticed on other pieces! good to know that I should have it pause for a second after Y movements if I need that not to happen.
Looking at the scale (those big blurry silver chunks are 0.01in / 0.254mm), center to center), I think I should set the X to be somewhere around 0.2mm (that’s the size of the X kerf) and maybe 0.05mm (or less?) for the Y.
I’ll try that out and run another test.
That estimate seems reasonable. Is probably rerun that exact same test with those values you’re estimating entered into the M425 compensation and see if it’s better. Tweak as necessary.
The wobble is kinda weird, I’ve heard of issues with the bearings inside the modules, that could be that. They are a weird size that I haven’t seen sold stateside, only on alibaba. Probably best to ignore that lol.
I got it to about X0.12 Y0.03 and the lines are so close as to only be separable under the 100x I’m using. Still some weird deformation on the side of the circle, and some gaps between the ends of the lines (not in the backlash direction, but along the axis of travel), so I’m making another batch to run at 100mm/min (these were done at 1000) to see if turning down acceleration would help. If it does then I’ll go figure out how to do that. Do you know in what instances (if any) M425 will reset and need to be input again?
The wobble in this case seems to be from my bed’s Y movements, or perhaps the entire shelf it’s sitting on. Not impossible that the movement is causing the whole thing to jiggle slightly. I wonder if the accelerometer on my phone would be able to pick it up…
