Newbie- new setup, accuracy issues

[DogWood]

Hi,

Just started milling on my newly cnc'd X2, I have a small piece that I created in GCAM (using EMC to drive the steppers).

What I'm seeing is on the y dimension the part is .020" to small and the X is .010" to large.

I have what I believe is very good repeatability, the cuts are all nice and lined up and I can run the program again with very similar results.

I'm using the stock screws at this point and have taken account of the backlash in EMC, could my dimensional discrepancy be from a incorrect backlash setting?

on the Y I'm trying to take .0105" off each side of an .375" wide piece of stock, it looks like the front cut is deeper than the back cut.
and on the X its the opposite, left (start) is shallower, right is deeper

any ideas?

[acondit]

First question, EMC? or EMC2?
How are you driving the z-axis?

Trying to have EMC2 control the Z-axis backlash can be very tricky on an X2 because it wants to pull the head down into the material when cutting (at least on a manual machine).

I would first just try to measure the movement with a dial indicator in both directions (x and y) one at a time. Put a rod in a collet. Zero the dial indicator against the rod, move away from the dial indicator 0.500 inch and then back towards it 0.500" (program control) and re-zero the dial indicator, then move towards the dial indicator another 0.500" note the reading on the dial indicator. Then under program control move back 0.500 inch towards the original position note the dial indicator reading. That will give you a feel for whether it is moving the commanded distances or not.

Alan

[DogWood]

EMC2 , Z- axis is is driven thru a pulley setup on the micro-feed, spring support removed. (a la chukkie)

I'll give the indicator idea a try, thanks Alan.

[DogWood]

ran a few tests tonight, sure enough had too much backlash on the X and not enough on the Y, ran a part through and it was very very close to what I wanted, within .002" X and Y

thanks for the help

[sansbury]

I've found that software backlash compensation usually causes more harm than good, at least until you've really got the machine figured out.