Lathe refrence frame shift?

HI guys I have been having some issues with my lathe and its z axis. during long runs with the bar feeder, the z axis reference frame will creep either toward or away from the headstock. After around 100 parts it is as much as .020 off. I checked the belts and servo pulleys and everything is tight and in good condition. This machine uses Mesa 7i33, 7i37 and 5i23 cards running yaskawa sigma5 servo system. It almost like the machine is missing some encoder counts or something? Im not sure where to start on a diagnosis on this issue so any help would be appreciated.

thanks.

Sounds like a probable electrical noise issue. Double check your grounding and cable routing (especially anything to do with the encoders). Are you using shielded cable for the encoders? Are the encoder cable run seperately from power cables? Can you try looking at the encoder pulses with a real scope. Hal scope probably will probably not be able to see the problem since it only samples those signals once/servo-thread cycle.

Are you using differential encoder signaling? ( if possible you should always do this )

One setup option to try is enabling the encoder digital filter, and if this is already enabled,
lowering the filter sample rate

The encoder cables are Yaskawa cables so I imagine they are sheilded. The servo drives send the signal to linux on foil sheilded cables as well. the encoder cables do travel in a steel tube to the control with a 120v ac line for my pc power though. I don't know if that would be enough to cause issues. The encoders are set up as rs422 differential. Is the digital filter built in to the anything IO card? or is it in linux?

Here is a photo of the back on the control.

The filter and sample frequency setting are part of the Hostmot2 firmware

The filter bit is per channel. When set, the encoder input requires 15 clocks of the sample frequency
to be recognized. When clear, only 3 clocks are needed.

To set, you would have something like
setp hm2_5i23.0.encoder.00.filter true (for encoder 0)
or
setp hm2_5i23.0.encoder.00.filter false (for encoder 0)
to clear

To set the sample frequency, you would do something like
setp hm2_5i23.0.encoder.sample-frequency 10000000
( for 10 MHz )

To know how to set these appropriately you need to know the maximum possibe encoder count rate

I looked inside my hal file and the filter is enabled, however it does not specify a frequency anywhere, so most likely the default frequency is being used. The encoder is 2048 pulses per rev by default. so should I set my filter resolution to 4x max encoder pulses per second?

What is the maximum RPM?
Is that 2048 PPR ( 8192 counts/turn? )

max motor speed is 3000 rpm and it is a quad b so 8192 counts per revolution. However the ppr from the servo drive is user selectable. It is a 20 bit encoder so technically it can provide much resolution, however I don't think that would be the cause of my problems.

Can you verify:
1. That the 7I33 is set for differential inputs (W1,W2,W3,W4 "UP")
2. That you have differential outputs (measure _across_ the A, /A pins B,/B pins
you should get a minimum absolute voltage of 2V across the inputs when static
that is less than -2V in one state and greater than +2 volts in the other state)
3. Since you have 8192 counts/turn this is about 400 KHz maximum count rate at 3000 RPM
this is roughly 100 KHz square waves to ~5 usec high and low times so with the filter on with 50%
timing margin thsi means the encoder sampling frequency can be dropped to 6 MHz. If you dont show
the encoder sampling frequency this means you should probably update to LinuxCNC 2.7

On the 7I33, all the jumpers are up except 4 since that is my mpg imput. I am measuring +/- 2.6 volts on the encoder channels. The voltage is not steady though, as I pulse the mpg I notice that the voltage oscillates a bit to get up to 2.6. sometimes it changes state but hovers around 1 volt. In any case it is not a crisp state change, and I suspect this is where my error lies. Problems is worse when I am using the .0001 jog increment. Not sure why it is doing this though. I am on linux 2.5 right now.