1/32 +/- errors mach 3

[MarkRH]

Hope someone knows the answer to this, my mechmate is direct drive with leadshine type drives.
This is over more than 50 sheets of plywood with many parts, it seemed to be accurate drilling 1/8" holes on four corners of a sheet of 5' x 9' mdf as I initially tested it, but with more lines of code its not as accurate.
I am getting errors in x and y dimensions +/- .04", even running the same files several times and it can happen between different parts in same file.
What ive tried:

(no slipping pinions they are keyed)

1. changing dip switches to higher resolution
2. changed the computer from api to standard
3. changed settings in mach to 10 then 15 pulse and direction
all wiring is shielded and there is no missed step clunks, and the parts are not slipping.

I'm thinking its my windows xp computer and mach 3, how much accuracy should be expected from this combination?

[jhiggins7]

From your description, it sounds like you are missing steps (or adding steps). Is the error always in the same direction (i.e. the axis is moving too far or not far enough). Have you tried the the "axis calibration" under the "Settings Tab" in Mach3?

I'm also using Mach3 on an XP computer. I just cut nearly 20 parts at a run time of 2 hours each requiring absolute repeatability and the system performed flawlessly.

I had a problem that turned out to be the notorious "loose grub screw" problem, but otherwise the MechMate performed well.

Since you seem to have checked out the "mechanical" sources of the problem, I tend to agree with you that it's something to do with the "electronics". Software doesn't tend to introduce "random" errors. So, it sounds more like "noise" or an intermittent signal (like a cable with a broken wire).

[MarkRH]

Hi John,
Yes I did the axis calibration several times, and after each times I changed dip switch settings and still has the same errors that are random in x and y, if I cut a 12 x 12 square it can come out 12.04" or 11.96" cutting from the same file. It seems to be worse with larger files, sometimes if its cutting .75" deep in 3 passes the last pass will be as much as 1/16" off leaving a small ledge only on 1 side(this has happened only twice).
I tighted the spring tension to second hole, still does it.
The wiring is all new cnc rated shielded and grounded 1 side.
All power supplies are in a separate box as the drives and board, I am using a vfd to control the spindle, that is in another box 7ft away and the wire is new shielded vfd wire thats ran with dust collection hose on boom completely separate. One thing I did not do is put the filters on vfd wiring, I did add ferrites to the line, any way to check for accuracy without the vfd running to make a cut? not sure a spring loaded pen or pencil would be that accurate.
Ive also tried reducing speeds, im cutting between 120-190 ipm.

[lonestaral]

I put in 2 separate earth cables.

One from the VFD, and another from machine frame
and computer.
I throw a bucket of water on the ground spikes now and again.
Static due to dust can be a problem.

[jhiggins7]

Does your MechMate ever cut a simple file without errors, like your 12 x 12? Is your Axis Calibration repeatable? That is, after you perform the test and then repeat it, is the movement accurate? I assume you perform the Axis Calibration with the VFD off and the Dust Collector off. Do you have any other potentially noisy devices or machinery in the shop? I once had a UPS (Uninterruptible Power Supply) that generated so much noise that my X10 technology did not work.

Since you've made a lot of changes, you may need to begin to reverse the changes until you get back to the system that worked. I know this is tough. But a random error is always the hardest to figure out. If you can pinpoint when the error began to appear, then you can document (for yourself) the configuration that was error free. Once you've done that, begin to reverse the changes systematically. The KISS principle applies.

[MarkRH]

Al now that you mention grounding, went back and read Geralds thread on proper grounding and realized didnt have it done correctly, fixing this now.
John, yes calibration is good and repeatable with no vfd, no dust collector on, also have 4 whole house vacuum motors for the table that also are off.
Thanks for your help, will update

[jhiggins7]

Hope the grounding fix works.

[MarkRH]

redoing the grounding, it seems to have helped a lot. I still have to drive a couple ground rods for this sub panel the is mechmate running from (there weren't any).
John, the errors were both + and -, and not on all parts in same file. For instance, a cabinet side that was drawn as 12" x 30", id get 12.04 x 30.04, on the same sheet id also get 11.96 x 29.96. This is reading from a tape measure estimating 1/32".
Just cut a few small pieces, the error has seemed to greatly decrease from .04" to .02.
Cutting a 6" circle its within .02, x measured 5.99, y measured 5.97 average.

One thing i did notice and not sure if its right, if you hold your fingers on the y motor plates while moving, they seem to very slightly move up and down, x does it less, I had the pinions bored and broached by a friend but he did it with a pocket toolpath and measuring with calipers I couldnt see a variation. But they seem to ride up and down just a small amount as they are moving.
Thanks

[swatkins]

Subpanels are by code NOT to have ground rods!

[MarkRH]

Detached building or structure they are.

[DocTanner]

You may want to check that the gears at firmly attached to the stepper.

[MarkRH]

Hello Don, yes the steppers are keyed so theres no slipping. I cut a few parts with dados and it seems to be within 1/2 of 1 mm now from part to part within the same file, in either direction. It definately got better redoing some of th3 grounding on the table.

[jhiggins7]

Improvement, good!