G2/G3 resolution

[gnrules]

Hi Tom,
i'm trying to smooth as possible a round path of 300mm diameter.
we tried different approaches using:

-corner tolerance 0.01mm
-Beak angle: 20 deg
- facet angle:0.1 deg
-collinear tolerance: 0

both with G2/G3 and tessellated path (chordal deviation 0.0005mm) we notice some sort o facets on the outer surface of the circle.
no change with number of passes, finishing, tool speed and feed (at least not significant).


I thought was the lnear segments but also the G2 gives a similar result.
my question is: how does kflop process G2/G3 arcs? I thought it calculates a sort of time discretization adding a point every n servo cycles, but maybe i'm wrong. I was expecting a very smooth swurface from G2, but probably there are some params that influences the actual number of points generated.

Thank you very much

Best regards

Giancarlo

[gnrules]

Forgot to mention we use KLP=0.002 filter and the sound of the mill is quiet and smooth, so no vibrations are in place (also changing tool loads with no effects is a proof).

[TomKerekes]

Hi Giancarlo,

how does kflop process G2/G3 arcs? I thought it calculates a sort of time discretization adding a point every n servo cycles, but maybe i'm wrong. I was expecting a very smooth swurface from G2

It calculates points ever 90us servo cycle using 64-bit floating point sin function so from that perspective it should be very smooth.

What Version are you using?

What are your axes resolutions?

How much do you think the facets deviate from a theoretical arc?

Do you have servos?

I don't think the filtering would do anything but make it smoother with slightly smaller diameter.

You might capture the actual trajectory (Destinations) with a C Program to check for any anomaly. I think you did that a few years ago here.

[gnrules]

Hi Tom,
sorry for the late reply but here in Italy there was a long bank holiday and I was on a trip.
the machine has servos, glass scales 0.5um resolution and we are using last version (5.4.1 maybe).

we performed some investigation acquiring data both during cut and with dry runs and my suspect is a borderline stable control loop.
the trajectory reveals almost perfect commanded points but a stable oscillation of +- 0.01/0.015 mm around commanded position.
while it is ok for the geometry it is clearly visible as "facets" because moving at 1100 mm/min means a sine wave with 0.8mm period on surface and 0.02/0.03 mm peak to peak (chordal error is much less, about 0.5um).

we are changing our lead-lag stage to be less aggressive and see what happens, stated that similar machine with less aggressive loop does not exhibit such an oscillation.
I'll keep you up to date.

Thanks

Giancarlo

[TomKerekes]

Hi Giancarlo,

Thank's for the update. There isn't any rush on our end.

Does the oscillation show up on the Step Response Screen moving at 1100mm/min. I think that works out at 23Hz?

Good luck.

[gnrules]

Hi Tom,
I performed a lot of test and a strange behaviour is on.
I tried moving a lot my loop parameters putting filters on and off and lowering gains to minimum, but the oscillation (about 20Hz) is always there.
Obviously following error changes during those test, but the fundamental behaviour and oscillation is still there (even in step response screen).
The suspect is a resonance that the control enhances, but I can't figure out which kind. the machine is very stiff, roller guides, 60° bearings, 25mm ballscrew with 10mm pitch. Everything is preloaded and nothing let me think there is some low compliance or any backlash.
But the behaviour is still in the area of a resonance.
Probably is something we can try to address with a notch, but I'd like to understand where it comes form ad I have some stability concerns for the very low frequency.
We close the position loop on a glass scale on 0.5um with the servo drive controlled as current amplifier. maybe this enhances the problem.
Do you think that changing to velocity loop in the servo may help? we close the position loop anyway on the glass scale, but it should be more damped.

Thank you for your support.

Giancarlo

[TomKerekes]

Hi Giancarlo,

Absolutely that is likely to help a lot. I assume the Drives are connected to rotary encoders on the motors that can provide feedback on what the motors are doing that would not be available in the linear scale information. Take for example backlash. Say the servo is trying to correct a small error so the servo ramps up torque. At some point the torque overcomes friction and the motor begins to accelerate but because of backlash the linear scales do not indicate this. So the torque continues to ramp up and acceleration increases. Finally after the backlash is taken out the scales suddenly show the movement but too late to avoid an overshoot. Whereas with motor feedback the Drives can move the motor at a controlled velocity through the backlash.

With only linear scale feedback there are so many things in the control loop that it is very difficult to have a stable system. Dual feedback is almost always required to achieve good performance.