In my case, just like Gerry's reply, I was missing steps and pulled my hair out trying to find the culprit. Machine 2.5D all day,but would randomly skip steps on 3D stuff and you coud not hear when or where it happened. Playing with settings didn't do any good.
So I replaced the steppers with ones of  two and half times  the original torque rating, got rid of the backlash in X and Y axis by replacing  the bearings and adjusting the ball screw nuts, and life is good again. Now if i would only get off my lazy butt and redo the Z axis life would be really good on the mill.
RICH

I  have slowed the z axis way down and reversed the settings as someone else suggested.  It is much more accurate now.  I am making a gun stock.   Lots and lots of Z movement. It always worked O.K.  if I finished using the long axis X,  but when I finished using the Y axis going the short distance back and fourth across the material there is a lot more z movement

Thanks
JR

Hi,

I had a similar situation - Z axis would "drift away" after many moves.
Eg. after test case with 500 runs of subroutine that moves 10mm up+down, zero position would consistently be arround 4 mm to low. That makes arround 0.004 error per move (less than one microstep in my setup which is 0.0125mm).
Anyway, after I tried/checked all that was suggested in this thread, finaly changing the kernel speed from 35kHz to 25kHz made it work perfectly.
HTH

I am already at the 25kHz It is close but still after 25,000+ lines it is about .1in off not bad on wood I can fix most of that with sandpaper.

Thanks
JR

JR,

Do you get lost steps on any other axis or is it just the Z ?

Tweakie.

Just the z.

Hi,

While I see this post has been dormant I am hoping you can steer me in the correct direction. I am seeing occasionally a loss of Z, in the .060" range.

I Zero the machine, move the spindle to 2.00" above, check with a 1x2x3 block (on the 2" edge) and looks fine. Run a couple of paths ( generated with Fusion 360) and I see roughly .060" clearance between the 1x2x3 block after moving the spindle to 2.0". The Collet is tight using torque wrench set at 50lbs (1/4" bit).

Remember the DRO is showing 2.0" so if I rezero the z-axis the DRO should never see anything above 2.0". Correct? That is not the case I see a value of 2.057" ish just as I hit the zero operation. See pictures below.

Its an AVIDCNC like, lot of the part are from AVID, the exception is the controller. I am using a ESS from the Boardroom, ClearPath Nema34 servo's and various PS.  If I was suffering from noise I would expect to see it on all axis. I don't I only see this in the Z-axis. Spindle rack is AVID and the Spindle is also from AVID. The Rack-Pinion is from AVID also.

Thanks
Loren

PS. I have simulated this and haven't seen it 'fail' using both NCViewer and the fusion built in version. the code is very simple, just taking .01." off the side wall using a 2D contour.

Hi,

While I see this post has been dormant I am hoping you can steer me in the correct direction. I am seeing occasionally a loss of Z, in the .060" range.

I Zero the machine, move the spindle to 2.00" above, check with a 1x2x3 block (on the 2" edge) and looks fine. Run a couple of paths ( generated with Fusion 360) and I see roughly .060" clearance between the 1x2x3 block after moving the spindle to 2.0". The Collet is tight using torque wrench set at 50lbs (1/4" bit).

Remember the DRO is showing 2.0" so if I rezero the z-axis the DRO should never see anything above 2.0". Correct? That is not the case I see a value of 2.057" ish just as I hit the zero operation. See pictures below.

Its an AVIDCNC like, lot of the part are from AVID, the exception is the controller. I am using a ESS from the Boardroom, ClearPath Nema34 servo's and various PS.  If I was suffering from noise I would expect to see it on all axis. I don't, I only see this in the Z-axis. Spindle rack is AVID and the Spindle is also from AVID. The Rack-Pinion is from AVID also.

Thanks
Loren

PS. I have simulated this and haven't seen it 'fail' using both NCViewer and the fusion built in version. the code is very simple, just taking .010 .  off the side wall using a 2D contour. The main cut is a Pocket at .020in over and the Final Cut is Contour taking -.010in. Both paths has the Axial at 0.