Following error - Rotary Axis

Why STEPGEN_MAXACCEL not allowed to be more than 1500?

With this config for rotary axis everything more than 30000 for max acceleration and 37500 for stepgen max acceleration gives following error.
Configuration is done by pncconf stepgen A rotary axis
Servo motor driver for 1 turn needs 2500x4=10000 pulses
and has 2500rpm limit

Plotting the following error vs joint velocity in halscope may give a clue.

Do you have the DPLL enabled?

Have you tried a longer DPLL setting (say -100 instead of pncconfs default -50)

(following errors due to host jitter will show up mainly during acceleration)



 

I played around a bit with this and learned that  with such
extreme acceleration, the stepgen PID loop must be tuned
(with added FF2 and I term)

_BUT_ I wonder why you need such high acceleration.
30000 degrees/S^2 means you get to full speed (10000 degrees/sec)
= 1666.7 RPM in 1/3 of a second!

testing servo motor... with 30000 is good acceleration
but i feel that can go 3 times faster mechanically (has low inertia).
I have got a 7i48 if drive it with analog -10+10V can go faster ?
How much FF2 & I needs for 60000 ?

What is better in general for driving servo that supports step-dir analog speed and analog torque ?

Is the axis actually driving a chuck?

Analog will not change anything.

Why do you require such extreme acceleration?
It will lead to inaccuracies when cutting.
This is multiple Gs of acceleration at any significant
part radius (if this really is a normal rotary axis)

If you want to tune the axis, you need to setup halscope
with commanded joint velocity and following error, then tune
FF2 and I (FF2 will be in the 0.0002 region)  Too little  FF2
will cause a positive following error for acceleration in the
positive direction. Too high FF2 will cause a negative error
for acceleration in the positive direction.
'I' probably in the 50000 to 100000 region.

You will likely need to increase the following error limits
for tuning.

Here is a tuned A axis with 40000 degrees/S^2 accel and
10000 degrees/S velocity:



 


Note that the peak errors are about 1/10 of a step.