Servo driver SG-AS20 YUHAI or CTB

Yes, this motor/driver is complete garbage. Today it started making bad noises. Looks like a bad bearing

youtube.com/shorts/onTwy6WiKgk?feature=share

Feedback in linuxcnc does not work for this driver. The acceleration must be very low, otherwise the pid controller compensates the position. Can the PID regulator be set with a long delay? Currently trying with P=1 D=0 and I=0 . I just want linuxcnc to watch for the wrong position.And when I turn of-on the motor, the program will know what position the screw pair

It sounds the PID/gain in the amp is set too hard. is there any software to look and adjust directly on the amp. PID of 1 or 5 in linuxcnc is ok. the servo has his own control loop and should be the master in the control loop. Look for parameters like Control Loop gain and decrease the values. 11 kw is already a large servo that needs to be adjusted well. 
edit
next step would be to measure the frequency and compensate with filter

This is a 1kw servo motor. I have two of these. It's not from a bad setup. I already tried with another one. If I have an acceleration greater than 200mm/s, LinuxCNC adjusts and sends more steps. So when the servo motor stops, it turns back several 160-300 steps. This is with a 10mm screw and 2400 RPM

I want to add a delay when reading the encoder. In this way, linuxcnc will get the current speed of the motor and will not make a correction with the wrong speed of the PID control. What do you think? How can it be done?

Have you considered not closing the loop in linuxcnc at all and just let the driver handle the motor? In my opinion it is futile to try building a PID loop with delayed encoder feedback.

Aciera wrote:

In my opinion it is futile to try building a PID loop with delayed encoder feedback.

+1

Yes, but I like having a second line of defense against bad wiring or shock issues. The other advantage is that when the motors are turned off (mesa works and LinuxCNC works), LinuxCNC knows the exact coordinates.

Maybe LinuxCNC can show actual encoder position and not use it for feedback and PID.
Not sure how exactly as i never used that, but someone here on the forum mentioned it a while back.

Ok, but how hard is it to add a delay when reading the encoder? I think this is the best way for me.
Or should I add linear encoders. But for one axis it is a difficult task.

Adding delays to real-time, mission-critical, life-threatening, etc stuff is never good.
You can add linear scales, that should work perfectly, also consider adding encoders to ballscrews or with pulleys if using belts.
-
Use the hal scope and measure what the exact delay is, probably could lover the base period time to half or 1/4 and see if that is still usable.
Just make sure the E-stop works.