Another servo tuning

Spikiness zoomed in. Where does it come from?

motor is 877.7143 / mm and encoder is 2000 /mm 
linear scale is 20um grating and EXE box multiply it by 10.

What do I do to make those plots believable?
 

I thought it could be the " hochohmig / nicht hochohmig"   (how's my German hey) on the exe box so I have changed high impedance to low with no visible change.
Any idea how to troubleshoot those spikes?

Well I just discovered that he f-error spikes must come from linuxcnc / mesa since in open loop 
this is input into f-error calculation instead of encoder
net    x-pos-fb    <=    hm2_7i95.0.stepgen.00.position-fb
net    x-pos-fb    =>    joint.0.motor-pos-fb

and I am guessing it will be the other input into the f-error calculation so the motor pos cmd? I dont know the inside of the system so it is my guess only but process of elimination.

is some wrong mesa setting? or is it linux cnc setting? computer problems? 
don't tell me that I am the only one in the world with this problem?

I have some work for the machine at the moment so I had to go back to original settings and slow it down again.
It is set like servo drives are in position control and linux is set like this:

#*******************
#  AXIS X JOINT 0
#*******************

setp   pid.x.Pgain     [JOINT_0]P
setp   pid.x.Igain     [JOINT_0]I
setp   pid.x.Dgain     [JOINT_0]D
setp   pid.x.bias      [JOINT_0]BIAS
setp   pid.x.FF0       [JOINT_0]FF0
setp   pid.x.FF1       [JOINT_0]FF1
setp   pid.x.FF2       [JOINT_0]FF2
setp   pid.x.deadband  [JOINT_0]DEADBAND
setp   pid.x.maxoutput [JOINT_0]MAX_OUTPUT
setp   pid.x.error-previous-target true
# This setting is to limit bogus stepgen
# velocity corrections caused by position
# feedback sample time jitter.
setp   pid.x.maxerror 0.012700

net    x-index-enable    <=>    pid.x.index-enable
net    x-enable        =>    pid.x.enable
net    x-pos-cmd        =>    pid.x.command
net    x-pos-fb        =>    pid.x.feedback
net    x-output        <=    pid.x.output

# Step Gen signals/setup

setp    hm2_7i95.0.stepgen.00.dirsetup            [JOINT_0]DIRSETUP
setp    hm2_7i95.0.stepgen.00.dirhold            [JOINT_0]DIRHOLD
setp    hm2_7i95.0.stepgen.00.steplen            [JOINT_0]STEPLEN
setp    hm2_7i95.0.stepgen.00.stepspace            [JOINT_0]STEPSPACE
setp    hm2_7i95.0.stepgen.00.position-scale    [JOINT_0]STEP_SCALE
setp    hm2_7i95.0.stepgen.00.step_type            0
setp    hm2_7i95.0.stepgen.00.control-type        1                                #0 position/ 1 velocity
setp    hm2_7i95.0.stepgen.00.maxaccel            [JOINT_0]STEPGEN_MAXACCEL
setp    hm2_7i95.0.stepgen.00.maxvel            [JOINT_0]STEPGEN_MAXVEL

# ---closedloop stepper signals---

net    x-pos-cmd    <=    joint.0.motor-pos-cmd
net    x-vel-cmd    <=    joint.0.vel-cmd
net    x-output    =>    hm2_7i95.0.stepgen.00.velocity-cmd
net    x-pos-fb    =>    joint.0.motor-pos-fb
net    x-enable    <=    joint.0.amp-enable-out
net    x-enable    =>    hm2_7i95.0.stepgen.00.enable

# ---Encoder feedback signals/setup---

#setp    hm2_7i95.0.encoder.muxed-sample-frequency 2000000

setp    hm2_7i95.0.encoder.00.counter-mode            0
setp    hm2_7i95.0.encoder.00.filter                1
setp    hm2_7i95.0.encoder.00.index-invert            0
setp    hm2_7i95.0.encoder.00.index-mask            0
setp    hm2_7i95.0.encoder.00.index-mask-invert        0
setp    hm2_7i95.0.encoder.00.scale                    [JOINT_0]ENCODER_SCALE

net    x-pos-fb        <=    hm2_7i95.0.encoder.00.position
net    x-vel-fb        <=    hm2_7i95.0.encoder.00.velocity
net    x-index-enable    joint.0.index-enable    <=>    hm2_7i95.0.encoder.00.index-enable
net x-pos-rawcounts        <=  hm2_7i95.0.encoder.00.rawcounts

If I don't use linear scales as feedback it drifts (X axis to the left/ negative direction)