Servo jump error

Hi guys I am looking for help with an issue with the z axis on my linux control lathe. I first noticed it several years ago but it was very rare. Now it seems to happen more frequently. The issue is that sometimes during a facing cut, the z azis will jump in the - direction by less than my ferror, leaving a ridge in the part. The last part I made was 400 pieces and about 20 parts had the jump. After removing the motor belt I noticed the motor has a visible jump or oscillation while turning it in the.0001 increment with the mpg, however it appears to be much worse when turning in the minus direction. I have attached a link to a video of the twitch. I have also included a hal scope showing the motor command and a reference speed trace from the servo motor monitoring software during one of the jump events. It appears that for some reason linux is telling the motor to reverse slightly. It is not consistent and I have tried a lot of different pid values and still get the twitch. Any help would be much appreciated.

Sounds a bit like an encoder problem

Can you plot the commanded position and feedback position?

Here is the motor command in red and the motor feedback in blue, green is the ferror. The output seem a bit noisy to me.

I cant tell from the photo but what is the time/division what amount of time
is the acceleration portion of the move?

X scale is set at 200 milliseconds per division in the photo.

it does sort of look like normal acceleration following errors
but I would need a better plot (say 10 ms/div) with velocity plotted
to verify

Here is a better sample. This is 3 clicks in one direction while in the.0001 increment. I noticed that the motor jumped on all 3 clicks. The second photo is a close up of one of the clicks.

Here are the hal and ini files. On the encoder what pins would you want to see? raw counts? The only reason I dont think it is the encoder is that the servo drive is not showing any errors. The Yaskawa sigma5 drive has on board monitoring and the first photo I show is from that software. The encoder signal to linux comes from the servo drive as shown below. On the photo I posted the reference speed incoming is going across the zero axis during on click in .0001 increment. I think this shows that the motor is following the reference signal ok? Correct me if im wrong but I dont think the reference speed should cross the 0 axis.



thanks

can you try reducing the deadband?