[awerby]

Depending on how you were measuring with your dial indicator. If it was loaded on one side of the threads when you made your original measurement, and on the other side when you made your final one, that would mean you've got ,011" backlash in X and .022" in Y, which isn't impossible to believe. If it cut your calendar correctly and each path was directly on top of the other, then that seems like the most likely explanation to me.

You still need to figure out what's causing the problem in Z, though. Have tried reducing your velocity and/or acceleration and running the test part again?

Andrew Werby
ComputerSculpture.com — Home Page for Discount Hardware & Software

All day while waiting to get home and run some tests I started thinking just how many things I've changed and all the settings I've tried. I figured it was time to go through everything and reset everything to what a 'default standard' config would be, ensure everthing was set per the manuals and even double check the wiring while I was at it.

Well, I came home and figured I'd set up a dial indicator and measure the Y axis (which was the Z axis logic/driver path. Well, it was off.

I have to admit here that I never did take any precise measurements of the X or Y travels. The progression went like this.

I got the machine set up and wanted a job to give it a workout. I chose the Aztec calendar floating around this site.



I set up a V-carve toolpath and loaded up some MDF for cutting. That's when I noticed the Z was changing. Over the course of a couple of weeks I continued to check/tweak some things and re-run the job after re-referencing to my home switches. Many, many times I turned everything off then came back later and turned it on, re-zeroed and re-cut the job with the workpiece never having been moved. I marvellled at how repeatable the switches were as other than the Z everything recut at the correct position. The jobs were run for various lengths of time never quite finishing before I would kill it.

This is with a design with lots of circles and sub-designs in circular arrays. The recuts along the X and Y were always right on so I never checked them.
So I don't know if things have changed or if this was always there. Frankly I don't see how it could have rerun the job so perfectly if it was off that much and varied with the job run time, but I don't KNOW that.

Anyhow, that's the story. Now for todays results.

I set up a dial indicator on each axis and referenced it. Then for each axis I ran a file that moved that one axis 1 inch and then returned. I repeated this cycle 50 times. Then moved it back to the reference point and took a reading.
Each test cycle was run twice on each axis and each run was within a thousandth of the each other.

Y Axis: 10,11 thousandths in the negative direction after run
X Axis: 21,22 thousandths in the positive direction after run

Here's the surprising one

Z axis: 0,0 after each run.

So, I'm not sure what this tells me but I'm thinking I do need to do a reset to baseline and proceed from there.

Any thoughts?

BobL.