Loss of Y position on fast jogging.

[AvB]

I don't know that this is a UCCNC issue, but it seems the best place to post it.

I switched from Mach3 to UCCNC a few months ago. We had to make some hardware changes to get it to work (notably changing the spindle control to be direct from the ETH-300 analog port) so there was a bit of rewiring, but it was all done well.
I also installed Gerry's 2017 screenset.

When I first got it set up, I did a small sign and found that the finish cut didn't line up with the roughing. Grr, I thought I'd made a dumb mistake. I've had the same issue on other small projects and kept blaming myself. I cut a really complex model aircraft wing mould and was really upset that again, my finish cuts didn't line up with my roughing cut. So today with new material and new toolpaths I began re-cutting the wing mould and was very careful with checking and referencing zero and soon realized I had the same issue again - it lost the Y position by about 2mm. So essentially I am in a holding pattern unable to use the machine until I work out what's the problem.

After quite a lot of testing I've found that the problem is that the Y axis loses position after jogging. I can see the error when I homethe axes - as the numbers scroll back at homing, the X and Z axes scroll back to perhaps -0.2 before coming back over the switch, but after jogging back and forth (or using the Park button) the Y will scroll as far back as -2mm when it's homing. So I can see it that way, but to verify it I also drilled a reference at X0Y0 zero hole and ran a big roughing toolpath, and when I returned the position to X0Y0, the cutter dropped perfectly into the hole showing that it had not lost position. BUT ... if I jog back and forth at full speed, or use the Park button, then return to X0Y0 I find that the cutter is no longer lined up with the reference hole, so it's lost position.

So ... does anyone have any ideas where we should start looking? The machine has servos and Gecko drives.

[AvB]

Update ... it does seem to be losing the X position too, though to a lesser degree than Y. I tried reducing the axis velocities for X and Y to lower than they used to be set in Mach 3, and it did not help.

[ger21]

If I recall, you're machine is using servos, not steppers? I think the only way you could lose position with servos is either mechanical, like a loose coupler? Or electrical. Either the drive is not seeing all of the steps for some reason, or noise is making them see extra steps?
But the fact that it only happens while jogging kind of throws that out the window.

What if you only jog slowly? Does it still lose position?

[Battwell]

also- half the acceleration and try again
are you working in inch or metric?
what are your steps per? and maximum speed set to? lets see if pulse stream is too fast for your drives.
do your servo drives support gearing?

[ger21]

Good point. If you have the kernel speed at 400Khz, try changing it to 200Khz.

[AvB]

Thanks for the tips gentlemen, much appreciated.

Yes, the machine has servos. Units are metric. Kernel speed is currently at 200. Communication buffer at 0.1 sec. Steps per is around 1640 for all axes. Velocity on X and Y are at 3500. Acceleration is at 300 for X and Y.

The Verify function macro is now working so I can measure the extent of the problem after a process. The most spectacular demonstration of the issue is when I use the Park function. I have Park 1 set fairly close to the home position. I only have to send the machine out to a job zero at say X500, Y200 on machine coords, using MDI (so I don't jog by hand). Then I press Park1 once, and it travels back obliquely back to the park position, then the Verify macro will show that Y is out by about 0.7 and X by about 0.2mm when it homes.

I also just ran a toolpath for a few minutes - a fairly jerky path with lots of changes of direction etc. I didn't jog by hand. When I verified it, Y was out by 0.2, and X by about 0.04. So I can't say that it only happens when I jog ... but jogging loss much more quickly than running a toolpath, and using Park is the worst.

I'll try reducing speeds, jog speeds, and acceleration.

[AvB]

Some test results and further findings.

I did a simple test routine 1) Home all, 2) travel to job zero using MDI (job zero is about X500, Y200 in machine coords). Then press Park 1 to send it back near the home position. Then run the M925 verify macro and check the errors.

VELOCITY:
At XY axis velocities of 3500, the errors were approx X-0.14, Y -0.48.
At velocities of 1500, the errors were approx X-0.06, Y -0.25, so the low speeds did more than halve the error, but it's still huge.
I then re-set velocities to 3500.

ACCELERATION: The original errors of X-0.14, Y -0.48 were with accel at 300 for X and Y.
At accels of 100 for X and Y, the errors were approx X-0.18, Y -0.57, which is significantly worse!

A couple of other points:
1) It seems to be intermittent with manual jogging. I did some tests running back and forth, or rectangles, or oblique (X and Y concurrently) jogging moves where I went back and forth 10 times, and sometimes there was no loss of position. BUT if I use Park, it will ALWAYS lose position badly. Park seems to be quite a fast move.

2) Looking at the diagnostics screen, the 017 LED is blinking red erratically. I don't know much about the setup but looking at the ports and pins, Charge Pump 1 is on Pin 17 on my Port 2. Is this normal?

HELP!!! haha - thanks for the input guys, it's much appreciated as I am totally stuck here.

[cncdrive]

Can you monitor the step/dir counts and the encoder counts with your servo drives?
If it has such feature then you can verify where does the system losses the position, in the encoder or in the step/dir lines.
I think it will be probably a noise problem, bad grounding and/or EMI noise effecting your drives.

[AvB]

I don't know how to do that, or whether it can be done. Where would I look? Sorry I'm no expert on the system.

[AvB]

I opened up the box on the machine to have a look for earthing problems etc. I'm not an electronics person and my engineer friend did the installation of the ETH-300 and helped me to get it set up.

Looking at the ETH-300, I note that the top section of the board, where the ethernet cable plugs in, has copper rings which look like the mount points, but we've mounted it differently. Should the top part of the board be earthed via those rings, or is this OK?