I have been working on my g0704 to make it as accurate as possible.

   Last week I adjusted the gibs with a dial indicator for lateral play in the table and was achieving 331ipm at 45k kernel speed with no positional loss whatsoever. I used the warm up wizard and created a file for each axis at a time (only using 1 axis at a time during the tests). I was above 200 ipm on all axis with no loss of steps.

   Those speeds were wonderful for the ego but not likely for anything I am going to do. I backed all axis down to 80 ipm max and retested many cycles to confirm no loss. Everything was great.

   Then I loaded a file (a stepper mount) which had some profiling, pocketing, drilling operations going on. Nothing complicated or fast. I was running between 8-30 ipm. However lifting out of 1 of the slots and moving to the next slot created a positional loss multiple times over. I reduced my speeds in half but it still loses position.

   All axis are still capable of flying and are accurate.

    Somehow I wonder if the computer is overloaded or not capable of the necessary pulse stream during 3 axis simultaneous moves. It just seems so odd. It almost as if the acceleration rate just went thru the roof for a moment.

   Reviewing the GCode there is no F value above 30 ipm. Also ran the wizards and cut a few pockets and circles with no issues and they measured to within .005 that I could tell on X and Y.

Why are you running the kernel at 45kHz?  You should use the lowest kernel speed you can, and maintain your rapid speed.  There is *no* advantage to running the kernel faster than necessary to enable the maximum pulse rate you'll actually use.

What acceleration are you using?  Setting acceleration too high is a good way to cause lost steps.  You should determine what acceleration *appears* to give the best performance, then back off 30-50%, to provide plenty of margin.  If you set it to the maximum that *appears* to work it will NOT be reliable.

Regards,
Ray L.

Ray,
  I had moved my kernel speed down to 25 after I realized it was set to 45 during my testing for lost steps. There were none. I was running a velocity (at the 45 kernel speed) of 331 and an accel of 20.  I then lowered my kernel speed and never sought a max speed. I set my velocity to 80 accel 20 for all subsequent use. Then today during the running of the file I heard it squeal and it loses position in the X. I lifted the Z, rezeroed it above the workpiece and reran the file. Sure enough it repeats the loss same place. I lowered the velocity to 45. Still does it. I also noticed it during an attempt at a chamfer pass on a small hole. Seems like if there is a bit of interpolation it goes haywire.

Is there a way I can watch the load on the CPU while I am running the file and see if it is overwhelmed? It is a different PC than I have ever used. Dell Optiplex G120 IIRC. It seemed to pass the driver tests and I went thru the optimization steps when setting it up. Minimal software on that computer and no Internet connection nor am I running other programs simultaneously.

Reduce acceleration, and see if that helps.  It you were able to run 331 IPM, then CPU load is not the problem.  Acceleration is MUCH more likely to cause stalling.  Or, there may be something running that is causing jitter on the step pulses, but I doubt it.  If that were the case, you would've never reached 331 IPM.

Regards,
Ray L.

Open the task manager to the Performance tab and you can watch the CPU usage.

How about the motor/drive power supply?  I've found over the years that when you can achieve high speeds on one axis at a time during testing, but then it fails when the axes move simultaneously, many times it's that the power supply can't provide enough power when all three are running.

PMDX 126, 5056D and a brand new 48V PS.

   So far the accel solved the problem. I reduced it to 7 and was able to put my velocity back into the 80's without issue. Audibly I can barely tell a difference if at all. I dont see the stuttering or loss nor do I hear it. My machining rates are fine. The issue was only occurring right in the corners of a few slots and holes that were very close to the diameter of the tool. All appears to be fine. Proof in MDF looked great.
                                                                                                        Thanks