Tunning PID Loop, encoder on the linear axis, not on the motor shaft

Well once again here, ready to continue learning, I have obtained some ethercat modules ek1100, el4032 and el5151, el1008 and el2024, I have a Mecal sw453 machine, I continue with mitre saws haha, until now I had carried out all the tests with (H-Bridge PWM) + (DC Motor with encoder coupled directly to the shaft), for the case in question a servo amplifier is used which I have already talked about here in the forum at another time (Maxifet AF 60 10/20) it takes a control signal +/- 10V and also requires a tachometer the motor power supply is 60V, the motor is coupled to a reducer which could verify the transmission ratio, the encoder is 250 ppr and when connecting it to the 5151 module I was able to make some measurements such as the encoder-scale is 10 since for 2000mm I got 20,000 counts.After this hardware introduction, I can add that I've run tests tuning the control loop, and although I've managed to move the axis, I frequently get tracking errors. Some time ago, I asked here on the forum if it was possible to configure this without the encoder on the motor shaft. Based on some exchanges with Tommylight (thanks a lot for everything), he told me that a control loop with this configuration was viable and that Linuxcnc didn't care where the encoder was. Obviously, everything has its difficulties. I've run some tests, including: I increased the servo-thread frequency to 2 kHz. I tried tuning the PID loop by adding P with I and D set to zero, but the system oscillates with gain P=1. Finally, after much trial and error, I arrived at P=6, I=3, :4-6, FF1=1, which produces movement, but at the moment of braking, it begins to oscillate. If anyone can contribute ideas on how to improve the loop response for this configuration they would be very welcome,

This is just a stab in the dark, but try
P=0.2 up to 0.5
D=0.01 up to 0.05
FF1=0
Everything else PID related to 0 except some deadband, like 0.001
-
This is not an assumption, this is the only thing that properly works for my Mazak with Mitsubishi drives
And, yes i did a lot of head scratching, it did not help

Hi Tommy, thanks for your quick response. These are my ini and hal files if you'd like to take a look.

In this case the encoder resolution is 10 counts/mm (0.1mm) so I use a deadband = 0.05, This eliminated the oscillations that occurred simply by enabling LCNC.

Hello Tommylight, your approach is very accurate. Starting from the values ​​you propose, I was able to make relatively smooth movements. I would only highlight the need to lower the D = 0.001-0.005 a little more, since the motor emits vibration with the proposed range of values. Regards and thanks.