[cyclestart]

At first I thought the screw couplers were loose but that wasn't it. The machine is a mini-mill with a g540. Reliable up until today. The code being run was CAM generated. Large files with many small moves. The commanded movements were mm to 4 decimal places, well beyond the capabilities or resolution of this machine. My normal programs are dimensioned in inches to 3 decimal places.

So the question:
Could the precision used in the gcode be the cause of the loss of position ? As an accumulative effect of very small moves and a very large file? One axis was out close to 3/16".

[universalfab]

At first I thought the screw couplers were loose but that wasn't it. The machine is a mini-mill with a g540. Reliable up until today. The code being run was CAM generated. Large files with many small moves. The commanded movements were mm to 4 decimal places, well beyond the capabilities or resolution of this machine. My normal programs are dimensioned in inches to 3 decimal places.

So the question:
Could the precision used in the gcode be the cause of the loss of position ? As an accumulative effect of very small moves and a very large file? One axis was out close to 3/16".

Nope your wrong, umm, have you had a power failure or your power shut off? I've had this problem one time, and the problem is your machines motor parameters. I almost guarantee you thats your problem, I've asked people on here when I had the problem and nobody knew what it was. It really pissed me off with all these know it alls on here and everybody was wrong and I figured it out myself. So yeah you go to your machine parameters and set up a dial indicator and re-calibrate your machine and you should be up and ready to go. Let me know what you come up with.

[cyclestart]

Arghh....another theory shot down in flames.

There were a whole lot of things plugged into that outlet. In addition I was using a a jigsaw, probably on the same line(?). Could it be a case of to much power demand ?

I'll check the setup. Using stepconf is good entertainment. Tommorow....it's getting late here.

[universalfab]

Is that a stepper or servo system?

[cyclestart]

Is that a stepper or servo system?

Steppers running on a gecko g540.

A little background:
It's a Sieg mini-mill fitted with 5 tpi ballscrews. I've owned it for a few years and I'm very familiar with it mechanically. It used to have Xylotex drives and has done a fair amount of 3d profiling with those drives. No problem.

A while back it was switched over to the gecko. Since then it.s only been used for drilling holes, cutting a few pockets, simple stuff and short programs. The step and direction timings are the ones provided by emc for these drives. The computer is the same one I've used for several years. All the electronics have ample cooling.

Right now the machine is running a loop, dropping to z -1 in in 0.100" increments and then returning to z 0 in the same increments. The loop is set to 1000 cycles. Nothing unexpected is happening. Next I'll try the same loop in 0.001" increments.

This is frustrating.

[lumberjack_jeff]

I was losing steps too. My problem was cpu load on my computer.

Next time you run a file, launch htop first. That will help you rule out that as a problem.

Don't run the gui "system monitor". The graph consumes huge cpu.

I changed video drivers and stopped using the backplot screen (I run the DRO instead)

[Zig]

Cyclestart,

Crikey mate... I hope the know alls did not rip You off while not providing assistance?

[cyclestart]

I changed video drivers and stopped using the backplot screen (I run the DRO instead)

I usually use the DRO when running. The 3 axis distance-to-go was a welcome addition to emc, can't imagine living without it now. You're right about the backplot adding to the load.

Anyway.....

The problem may be solved

The G540 settings in the stepconf drop down box are not sane. The tool was 0.150" out of position on the z at the end of a 3d program used to test.

Using a G540 stepconf file found on the emc wiki gave better results. 0.060" out of position on the z.

Adding another 1000 to the step timings brought that error down to 0.005". That's the approx backlash on the z, so it might be dead-on for steps.

The settings may still be a bit off optimal but I'm calling this one solved.

[cyclestart]

Cyclestart,

Crikey mate... I hope the know alls did not rip You off while not providing assistance?

I didn't ask for help from Gecko if that's what you mean. I respect their products. Wouldn't have bought one otherwise..