I have an interesting issue that I am having trouble diagnosing. I hope one of the kind people here can help.


System Specs.


Custom 4' by 8' CNC with HSD 12HP ATC spindle and 11"Z.
Very solid frame and gantry.
THK Linear bearings and 3/4" 12 Pitch 15.5 degree rack with 18TPI gears. Z is 16mm ball screw with 5mm lead.
Direct drive with Oriental Motor AR911AA3 Motor Driver combo (semi closed loop - will not lose position - if count go off from encoder it will correct the mis-step in the driver hardware or  throw an alarm and stop the system if the error is too great)
The gears are attached to the shaft via hub and are not susceptible to coming lose.
Smoothstepper USB with latest drivers.
Latest version of Mach
Control Computer is an i7 with 8GB ram and SSD HDD running Windows 7 Pro.
Design Software - ArtCAM Pro (Legit licensed version) 2013.

This machine recently ran with a different control system setup and ran without exhibiting the below issue.

If I am cutting circles, squares, rectangles or any non-relief part everything checks out fine. If I run a relief the part literally grows and goes slightly askew no matter which direction I cut it - x or y, it grows corresponding to the direction - in other words, if I am carving a number 1 for example, the base will be fine and eventually the top will skew to one side. If I change orientation and raster mill it along the Y as opposed to the X the exact same defect appears and looks the same as the former but the skew is always furthest from zero.

Another example. I did a bird with the wings spread. I cut and pasted the wing and mirrored it when I drew it so I know it to be exactly like the other. If I run it milling along the X the right wing (furthest from zero) will slant and be slightly larger and detail does not suffer. If I mill it along the Y the left (furthest from zero) will be slightly larger and also slanted. If I cut a box the same dimensions all lines are perfectly straight and to specified length.


Here is what I have checked/tried;

Inspected the part in the Mach MDI - Looks like the ArtCAM rendering.
Noise
Wiring
Driver logs (the drivers will store information about miscounts and other data issues and can be computer monitored)
Set screws
Backlash
I ran the part files on another machine with no problems.
Steps per for each axis - things are moving to where they are supposed to.
Different milling strategies
Alternate acceleration settings ranging from very low to average
Debouncing
Plasma Mode
Step/Direction Pulse Delay
Running the part file ridiculously slow
Running the part file ridiculously fast
Running the part file at optimum speeds/chip load
And several of the other obvious suspects
Milling at different points on the table


I have not tried reinstalling and reconfiguring Mach from scratch nor have I tried a new smoothsteper as I don't have one available at the moment.


All checked out. The drivers do not report any corrupt or missing data nor corrected mis-steps (basically means they moved as they were instructed) and the mechanics are in good order. There is nothing lose and everything is solid. The machine is capable of good acceleration without issue.

The Mach setup is pretty basic. Only changes are what were needed to function such as motor tuning, a few custom M codes and port/pin setup to match the configuration. Not an expert at Mach but not new either.


Could this possibly be a SmoothStepper issue or Mach Issue? I am doubting it is mechanical as everything was checked several times over.

I am at a loss here. Anyone have any idea as to what could be going on?

Thanks in advance for any suggestions.

I can't find a way to edit my post - I noticed a typo above - the rack is 14.5 Degree and not 15.5

I suspect that you are loosing steps do to either too high velocity OR too high excelleration.  and the longer you run the worse it gets position wise.  At the end of a run you should be able to Return to home with compare and see IF that is true.

Try reducing the Vel / accel settings in half and retest the cuts.

(;-) TP

Thank you so much for the reply.

I will give that another shot but I have tried acceleration settings all the way down to 5 IPM - Normal for this machine is 30 to 40. Also, if steps are missed and not correctable the driver will throw an alarm and Mach is set to stop on a driver alarm. (see http://catalog.orientalmotor.com/item/all-categories/r-series-closed-loop-stepping-motors/ar911aa-3?plpver=11&categid=100&prodid=3001095).

I have purposely overdriven the machine to test this. A missed step usually only results in a slight "jitter mark" from the position compensation built into the hardware. If it can't reposition for any reason it will simply trigger a custom input and halt the system.

You guessed right in that the longer it runs the worse the position gets but the whole thing scales up and skews and another odd thing is that if I return home and run the file again (as in over the previous cut passes) it perfectly duplicates the error. If I miss steps and for whatever reason my closed loop system doesn't detect the miss would my zero position not be altered as well?

Thanks again for the suggestion - I am trying again at even lower acceleration right now. If it works I will post back otherwise the issue is unresolved.

DO you know for sure the Drive system is working properly ? Have you stalled the drives to test???

Reduce the VEL in half as well.

Have you checked the screws for accuracy ??  OVer a LONG distance not just a short hop.

(;-) TP

Has the machine ALWAYS had this problem or is it a NEW problem ?

(;-) TP

The problem only recently occurred - last week it ran great. No settings were changed.

Oh - and yes - I have 2 backup motors and drivers. Tried them. Additionally the cables are factory made and well shielded - no signs of damage but I replaced x and y cables anyways. I have a digital slide DRO off my hop saw and used it to check travel to 50" on both x and y - it was spot on (at least to the 2 decimal places the DRO is capable of).

Does it Skew in one axis only ? one direction only ?? everytime or random??

(;-) TP

Like I said in my post - it skews away from zero on either axis but cuts simple vector paths very accurate. Only on reliefs do I have an issue.