[outsider787]

I recently converted a BF20, and am still using the stock screws. I measured the backlash in the x and y direction, and entered it in Mach3's backlash compensation screen.

All seems to work well until I started milling out some PCBs (milling the copper off the board). Since I'm working with small cutters, a lot of gcode is generated (getting close to a thousand lines) and there's a lot of fine movements of the mill as it's cutting away the copper.

I've noticed that if I run the same gcode a second time, everything that is milled is offset in both axis.

So I'm thinking that my backlash may be off by just a little bit, but after thousands of fine movements the backlash error gets multiplied?
Is that true?

If so, what's the recommended way to measure backlash? Cycle one axis back and forth a lot of times to get an average backlash?

[LacL]

Take at look at some of Hoss' videos on backlash compensation.

hossmachine - YouTube

Check to make sure you have no slack on the thrust bearings. Move each axis several times back and forth a set distance and make sure it always returns to the same pos to make sure you are not loosing steps.

If I'm not mistaken, stock screws have linearity errors so different sections may have slightly different backlash amount as the threads may not be evenly spaced along its length.

I'm also looking to make PCBs on my mill as soon as I solve a few tech issues of my own. What software are you using to generate the gcode from the gerber files?

[outsider787]

LacL,
Linearity errors would make sense, and could account for the fact that I don't come back to the inital zero location, even though I have decreased the speed of cutting considerably (so I don't think I'm loosing steps).

The most straight forward way I've come across is using CopperCAM. Very simple to use, but has some limitations. For straightforward isolation it gets the job done.
Import the gerber and excellor files from Eagle (or your fav package) and Coppercam exports gcode for both drilling and isolation.

I don't have an issue when making small boards (2"x2"). The problem became apparent when I was doing a larger 5"x3" board.

adrian.

[691175002]

If the error gets worse on longer programs, you are losing steps.

If I had to guess I'd say that you are most likely losing steps when the screws reverse direction. Try increasing the time it takes for the backlash compensation to take effect. If the movement is too sudden you will lose steps.

I don't have my copy of mach3 nearby so I forget exactly what its called, but hoss mentions it in his videos.