I built a gantry style mill.  I surfaced the spoilboard no problem.  I've tried a few times now to cut out my first project and here is what's happening.  I have it set to relief cut with a 1/4" bit and it's set to make multiple passes at an increasing depth of 1/8" so it would be three passes to get to it's final depth of 3/8".  It makes the first several lines no problem with the proper depth and then Mach seems to be adjusting the depth and before I know it it thinks 0 it's 1/4" below the actual work 0.  It progressively gets worse and I can't figure it out.  It's cutting an area 12 x 10 and it will cut out 12 x 3 of it on it's first pass before it starts going wrong.  I can't see anything in the G-Code that is adjusting what Z is to change where it thinks zero is.  Any ideas?  G-Code is attached.  When I run it in the simulation it looks perfect.

Hi and welcome to the forum.

Whatever program did you use to create that code ??

As you say It starts off aright then ends up all over the place as far a Z position is concerned - Try the Surface wizard within Mach, it will produce better code.

Tweakie.

Pic of the passes. File works. I will assume that what is shown was wanted. 
Got to go ....later,
RICH

    I thought we were still on the surfacing, didn't realise there was a picture in there.   

Tweakie.

I used ArtCam to create the code.  It worked great through sim but not when I run in through Mach 3.  I've run drilling patterns and they worked fine.  Can't explain why I seem to be losing steps at a certain point on my z-axis.

Maybe because your Z axis just dosen't have the motor to do the job? ie; torque and acceleration
RICH

My motors are Nema 23 425oz.  I ran some tests to try and figure it out.

The code was simple.
G0 z1.000
G0 Z1.125
repeated over and over

Test 1: 50 cycles of code, 1" per sec acceleration
Result 1: Z axis is 0.0320 lower than start

Test 2: 50 cycles of code, .5" per sec acceleration
Result 2: Z axis is 0.0318 lower than start

Test 3: 250 cycles of code, 1" per sec acceleration
Result 3: Z axis is 0.379 lower than start

Test 4: 250 cycles of code, .5" per sec acceleration
Result 4: Z axis is 0.380 lower than start

Test 5: 250 cycles of code, .25" per sec acceleration
Result 5: Z axis is 0.380 lower than start.

Seems acceleration has nothing to do with it. Are we back to voltages coming from the parallel port? Total z-axis weight is probably around 20lb with a Nema 23 425oz motor holding it.

I can only imagine how much worse it will be the longer I run the code for.

Spacecase,
Can you post a pic of your Z axis showing how it is driven and also your xml file?
How did you arrive at your velocity and deternine your steps per unit used in motor tuning for the Z axis?

RICH

Is the z axis actually moving .125"?

If yes then the steps per are correct.

If no then reset them so it will move .125".  this in no way affects the problem you have though, even if the steps/per are wrong the axis should still repeat.

Is the start position moving closer to the table?

If yes then its loosing steps when trying to lift the weight of the head, slow things down, reduce acceleration and velocity by half, and/or counter balance the head.

If no more like a breakout board problem or electrical interference, is the router spindle power cable next to the z axis motor wires.

yes the z-axis move .125 when asked to move .125.  It was calibrated to move accurately.  It is moving closer to the table so I did a few tests to determine if the spindle assembly was too heavy and it didn't seem to be.  It didn't matter if I ran the test at 1" per sec accel or .25" per sec accel, the distance off was withing a thousandth of an inch which I attribute more to my ability to read my dial indicator.  I'm putting in a PCI Parallel port today in hopes that it's a full 5v instead of the on-board parallel port which I'm going to test shortly but am thinking it's 3.3v.