Drift on servo system

Hi,
I have LinuxCNC on a milling machine. Servo + encoder + pulleys + ballscrews
The servos are driven by PWM on // port. The encoders are read by // port

Now I have a drift problem on X and Y axes. It seems that the problem is recent.
It seems that the problem is more important in case of circles

I suspect a sliding between encoder and ballscrew or a problem on reading the encoder.
But :
- I have placed some marks on belt/pulleys : all is OK
May be the problem is between encoder and motor ?
- linuxCNC doesn't display any error message like 'encoder read error'
The displacement is lower than it could be. that means that linuxCNC has counted too much pulses from encoder. That means that I have a noise problem which implies a quadrature error reading.

My question is :
Is somebody absolutely sure that linuxCNC displays a message in case of quadrature error on encoder ?

Thanks

At least on 2.6.0~pre I have, as I saw it yesterday before using a machine, and I was not sure about what it was and I searched for information.

The software encoder component does not have any error reporting, but the Hostmot2 hardware encoder counter does (Other hardware counters may have error reporting, but I am not familiar enough with them to know)

The actual error reporting in the hostmot2 driver was only added in master, and unfortunately has a bug so may have false positives (report error where none exists) on startup

So that is why I am getting those errors on my 5i25/7i85 on start up.

Is you drift a slow constant creep or is it errors after movement? Is it always in the same direction?

Checked you motor shaft couplers and pully hubs... for slippage as well as your encoders.

Yes its a bug in the master version of hostmot2

OK. Unfortunately I use the software encoder.

I have checked shaft couplers and pulleys. The only thing I can't check is the coupler between the encoder and the motor, but it is on the rear shaft of the motor. No torque at this place.
The feed speed and the charge of the CPU have no effect on the drift.
So I am probably facing to a noise problem. The 'false' pulses are not many, so the source of the problem will not be so easy to find out.
I will make some measurements on the machine at the end of january, I will report them.
Thanks