Improving precision?

Currently I have a working router CNC running on linear bearing and driven with steppers and ballscrews. With Mach3, SmoothStepper and a G540. To make a long as short as possible, my machine is only reliable to 1/100 inch, cheap ballscrews and nuts, backlash and probably other things as well.

Maybe it's worth mentioning, I have a dual drive X axis.

Yes, I know this is a LinuxCNC forum, I'm here because I'm exploring all possible solutions.

If I had a linear encoder either relative or absolute or whatever in fact such that could (even manually) drive my X axis to this or that position with a 95% confidence that the error is no more then 1/1000 inch or better, I'd be a happy man.

I make small wooden boxes using the V fold technique and I just need to make 3 or 4 V cuts, I could do this manually and as it is my box are looking like shit. Unfortunately for me I have a good eye and I don't like what I see.

The rest of my work is not so demanding, I could live with what I have.


Yves

I have not had very good luck with miter fold joints. They are difficult and fiddly to get right, and even then did not scale well for production. And this was using large heavy duty commercial cncs ($100k+ moving table fixed bridge Komo). A lot of our problems can be blamed on not using materials well suited for miter folding (surface skins weren't flexible enough and cracked easily.)
Some things I learned were, 1st using the right bit is crucial. A 91 degree bit with a slightly flat nose rather than a perfect point, will usually give better results than a perfect 90 degree V.
Use a material with a flexible skin (vinyl or paper...). Wood veneers usually don't work well, unless you apply tape or some kind of film to hold the joints together before you mill them, and even still you are going to get a splintery joint you'll end up sanding mostly off (and still looks like poo)
A light duty machine like most cheap or home built routers probably is not going to be stiff enough to mill to the consistent depths needed to do good folds. To try to compensate, mill the fold grooves in multiple steps and use a final clean-out cut that only removes a small amount of material (0.02 or so). And if your stepper machine is loosing ANY steps on the Z axis the whole process is going to be hopeless (a real and probable scenario with Mach's tendency for acceleration constraint violations)

Good Luck

I tried another method to calibrate my axis, one that I'm sure is independent of backlash. I'll also try make my gantry square. From there I'll do some more test and from there I'll decide.

Thank you very much for your input.
Yves