I am assembling a home-brew Mill using NEMA 34 steppers and have hit an interesting issue whilst trying to calibrate my X and Y axes.
I ran a simple test using a pen in the chuck to test the accuracy of rapid and interpolated moves using incremental G0 X10 and G1 X10 F30 steps respectively.
The result was rather surprising, as rapid moves appear to be accurate, but slow moves GAIN steps!
Pen Trace 1<http://www.kill-9.co.uk/tdc/frankenmill/ruler.jpg>
I ran another series of tests, changing one parameter at a time to see what influenced the problem.
Pen Trace 2<http://www.kill-9.co.uk/tdc/frankenmill/ruler2.jpg>
The test run parameters are as follows:
vel = max velocity under motor tuning for X & Y axes
accn = acceleration under motor tuning for X & Y axes
current = max stepper driver current
clock = Mach3 kernel speed
1. G1 F300 vel=2500mm/min accn=1000mm/s/s current=3.8A clock=60KHz
2. G1 F150 vel=2500mm/min accn=1000mm/s/s current=3.8A clock=60KHz
3. G1 F75 vel=2500mm/min accn=1000mm/s/s current=3.8A clock=60KHz
4. G1 F75 vel=2500mm/min accn=250mm/s/s current=3.8A clock=60KHz
5. G1 F75 vel=2500mm/min accn=50mm/s/s current=3.8A clock=60KHz
6. G1 F75 vel=2500mm/min accn=50mm/s/s current=4.58A clock=60KHz
7. G1 F75 vel=500mm/min accn=50mm/s/s current=4.58A clock=60KHz
8. G1 F75 vel=500mm/min accn=50mm/s/s current=4.58A clock=25KHz
9. G1 F75 vel=500mm/min accn=50mm/s/s current=3.8A clock=75KHz
10. As above, but loosened off the X axis gib strip
11. G1 F1000 vel=2500mm/min accn=50mm/s/s current=3.8A clock=75KHz
12. G1 F1000 vel=2500mm/min accn=1000mm/s/s current=3.8A clock=75KHz
So, it would seem that the table is losing calibration slightly at 300mm/min feed rate and is really bad by 75mm/min.
Lowering the acceleration marginally improves the error.
Increasing the driver current makes things somewhat worse. This is a pair of allegedly 9A drivers with 350NM steppers rated at 4.5A. The steppers barely get warm when run at 4.58A, but after five minutes at this current, one of the drivers went into thermal shutdown so I dropped back to 3.8A.
Lowering the kernel speed from 60kHz to 25kHz made the problem much worse. Pushing it up to 75kHz made little difference to 60kHz.
Loosening the gib strip on the X axis made things significantly worse.
Ramping the feed rate up to 1000mm/min brought the output back to 10mm steps again, but the reduced acceleration rounds the corners.
Running at 1000mm/min with an acceleration of 1000mm/s/s gives an output that is accurate to within the error bounds of the felt tip pen.
So, running fast makes it more accurate. Increasing the power or reducing the load makes it less accurate.
The steppers must be moving more than one microstep per step input at lower speeds, but I can't see how this can happen unless the inertia of the load was stronger than the holding torque of the stepper. If this were the case, I'd expect it to get worse with speed, not better.
Has anyone seen this before?
Thanks,
Terry