Not repeating with linear scales

[msmart]

I've gone so far down the rabbit hole...need assistance. Not sure if this is the right place but here goes. Lagun knee mill had Anilam Crusader control and I have retained drives, motors and the Z axis scale (have replaced the X&Y with imports 5080 cnts/in)
I am having an issue with accuracy/repeatability with the Y axis. when running a test program moving back and forth 750+ times, the position measured by 2 indicators fluctuates and is not consistent. The best I've gotten was .003" but at times as bad as over .040"...in both +/-

Kmotion is showing reaching position within 1-2 counts on each move.

I have:
-Manually tuned the Westamp drive as per instructions
-Inspected/adjusted mecanical components of Y axis, including cleaning and repacking the ballnut and making sure ways are getting lubed
-Attempted to best tune with Kmotion software
-Lost days of my life and hours of sleep

I am still not getting any consistent results, even if consistently inaccurate.

Any thoughts/ideas?
Thanks!

[TomKerekes]

Hi msmart,

Do you have KFLOP + Kanalog with Analog drives?

Do you have any feedback to the drives? ie rotary encoder?

Are you closing the position loop with the Linear Scales?

Do the Linear Scales have differential outputs connected to Kanalog?

I think you are saying you are moving back and forth some distance (what distance?) with two dial indicators measuring the position at each end of the move?

But basically the encoder's job is to track position. If starting at some known mechanical position with the encoder reading some value then moving around (it doesn't matter how or how poorly) and move back and the encoder reads the same as before but the mechanical position is different, then the encoder definitely has a problem. No matter how much slop in the ball nut, or friction, or poorly tuned and jerky the motion, the encoder should still be able to track the position.

A common problem is for noise to add or subtract encoder counts. Are your cables shielded? If so how? Are they near noisy motor cables? Noise counts tend to accumulate so eventually the error grows to such a large distance that it could never be a mechanical problem.

[msmart]

Hi Tom-

Yes-KFLOP + Kanalog with analog drives.

There is Tach feedback to the drives only.

Yes-closing the position loop to Kanalog with the linear scales.
Yes- moving back and forth a distance of 8" however the 2 indicators are both at the origination position of move (one dial and one digital)

The external cables are armored and shielded and in the control cabinet shielded with drain...and not really close to motor cables.

The scale is brand new as I thought the original one was problematic as I has a similar issue. However I did not test much, just replaced it. I don't seem to have this issue on the X, which has the same brand scale and the Z which has the original Anilam scale. I have been completely baffled because I believed that with linear scales there was absolute positioning.

Ahh... Here might be the problem...I don't think the scales are differential as the pinouts are as follows:

Pin 1 (red wire): 5V
Pin 2 (white wire): 0V
Pin 3 (yellow wire): A
Pin 4 (black wire): B
Pin 5 (blue wire): Z

Does this mean they are in fact single ended and could this be the problem? If so, how might I address it?

Thanks!

[TomKerekes]

Hi msmart,

Connecting single ended signals to differential inputs and just leaving the complimentary signals floating is most likely the problem. I'm always surprised it works at all. You might read this from our FAQs.

You could make the connections directly to KFLOP's single ended encoder inputs, but then you will not have the noise immunity that differential signals provide. If your cables are shielded properly and away from noise sources it may work ok.

But the better approach is to add differential drivers near your encoders. Here is another 3rd Party option:
https://cnc4pc.com/differential-line-driver.html

[msmart]

Hi Tom-

Thanks for your help! I think I'll try a test connecting directly to the Kflop and see how that works, but may end up purchasing the line drivers.

It is curious that I don't have this problem in X and Z...it seems to be only the Y. If they all were acting up, I would have suspected a problem with wiring and/or configuration problems.

I suppose i'll just add this to the long list of learning experiences

I'll follow up with my findings

Thanks!

[msmart]

Well that seemed to fix it! I ran the test program just once, but returned back to 0.0000, as it should.

I'm going to wire the X and Z directly to the Kflop and see how that works for me.

Thanks again!

[TomKerekes]

Very good. Thanks for taking the time to post back

[msmart]

Hi Tom-

One new development.
I'm now getting significant hunting on the Y (+/- .001" on the indicator and 1 count in Kmotion).
Can you steer me on this one?

Thanks

[TomKerekes]

Hi msmart,

It's common for servos to dither one or several counts. You might read this wiki article.

Based on your stated resolution 1 count of dither in KMotion should only be 0.0002" on your indicator. What might be happening is that you have some mechanical issues where the table is allowed to roll, pitch, yaw some amount. Because of this when the servo pushes and pulls on the axis even though the point measured by the encoder remains fairly stationary the point on the axis measured by the indicator moves significantly more.

[msmart]

This makes absolute sense. I do know there's some backlash on this axis.

Though in my desperation to get the axis to repeat, I was adjusting (loosening) the gib. I have snugged it back up and it is doing much better. It's still hunting, but maybe about +/- .0005...which is fine for now. I think I should take some time and revisit the tuning now that the scales are connected to Kflop.