big latency with KINCO FD-432-EA

Hi,

First a great thanks at linuxcnc.org and particular to Grotius for RTOS release

So, in my case I try to set up one axis


I tried in position mode and now in speed mode, it's working! movement and scale are ok
but the f-error is very too big, with at least 25 milisecond of "latency" between pos-cmd and actual-position (see capture screen).

And I don't understand why? I tried to improve gain and other parameters

(all servo's parameters are make with KincoServo+)

I guess this error is a clock fault ??
or other way ?



Actually this parameters are not clear for me in .xml file

        <master idx="0" appTimePeriod="1000000" refClockSyncCycles="5">  

and
        
        <dcConf assignActivate="300" sync0Cycle="*1" sync0Shift="-25000"/>  



I will be happy for any advice

hardware:
from en.kinco.cn/productdetail/cd3fd3zjzzzz05.html

AC SERVO MOTOR
KINCO SMC80S-0075-30MAK-3LSU

AC SERVO
KINCO FD-432-EA

Linux debian-CNC 4.19.0-11-rt-amd64 #1 SMP PREEMPT RT Debian 4.19.146-1 (2020-09-17) x86_64 GNU/Linux


 

Hello,

DC Sync and ref clock synccycles : forum.linuxcnc.org/ethercat/44113-etherc...l-file-configuration




there are a few points in your hal file that may cause this behaviour:

-place the line "addf x.pid.do-pid-calcs servo-thread" after the "addf motion-controller", with your current order pid will calc with the values from the wrong cycle. search in this forum for "execution order" i have explained this a few times.

-net x.s.pos-cmd => x.pid.command => axis.x.pos-cmd
your are using the wrong pos-cmd, "axis.x.pos-cmd" has to be joint.x.motor-pos-cmd . I believe this is the main reason for your delay.

Thank for your help,

I noticed your message on "execution order" and made some adjustments since then. ( this one )

I changed too axis.x.pos-cmd to joint.0.motor-pos-cmd

and increase gain for a best response
but always ~20ms of delay..

You can see on new screen capture (oscillo_3.png) where I have added joint.0.motor-pos-fb (green)

Why I have a delay between joint.0 and lcec.0 ?? and so why lcec.0.0.pos_error seems coming at the right moment ?? ( this will prove that  is not a clock / sync problem, no ?)

 

quickly:

My reference was not the right one: lcec.0.0.pos_error is about speed and not position. The good reference is joint.0.f-error