BACKLASH setting not working

Hi all, we have run into a bit of an issue. As we were trying to set up backlash compensation on our machine, subsequent tests showed that no changes were being made to to rate or lost movement due to backlash. We're running LinuxCNC 2.7.5. We are not running steppers, so the problem can't be the STEPGEN_MAXACCEL setting.

We would be very grateful if anybody could shed any light on this issue. Does anybody know what could be preventing the setting from taking effect? And what can we do to remedy it?

I'm a a little confused. One sentence you are talking about quantities of lost steps, And in the next sentence you say that you are not running steppers?

Could you clarify what hardware you are using?
Are you running servos? Are they position mode servos (step/dir commanded)? Are they closed loop (encoder feedback into Linuxcnc)?

Thank you for clarifying. No, we're not running steppers. What I meant to say was that there was no change in movement lost due to backlash, even after changing the setting in the INI.

Yes, we are running servos. They are closed loop from the servo to it's drive, but it is an open loop from the drive to LinuxCNC. And no, they are not position mode.

I appreciate the quick reply. I've edited the main post to reflect the proper wording of the issue.

How much backlash do you have? Could you upload you ini and hal files so we can have a peek and try to see what's wrong?

We have 3 axes, two of which have about .0015 inches backlash, and the other axis has roughly .009 inches.
The HAL and INI files are attached.

How are you testing the backlash?

My shot in the dark is your servo drives are not tuned well enough to notice a .002" movement...

backlash certainly works - I have just tested it recently.



first ring - backlash enabled - second ring backlash turned off.

Actually, you _are_ running your drives in position mode
Can you plot the commanded position when doing a reversal?

Is it possible the fast motion during reversals is somehow not accepted by your position mode drives interpolation?

I have no experience with an Ethercat system, but Linuxcnc is running totally open loop with no feedback at all as you have it now. So how can you have any idea if the drives are following the commands it receives?
Can you get any feedback from your system, for diagnostic purposes if nothing else? (a list of available hal pins might help find such things.

Good point. I suppose we have an unusual configuration.

We did our backlash test at incredibly low speed, stepping 0.001 at a time and using a touch probe against a piece of stock to see the difference in position between when the probe initially lights up, then backing off and seeing when the light goes out. So for that scenario at the least, I do not think it is a speed issue.

Somewhat of a side note, I believe we have a following error value coming back from the drive. Can we connect that to a pin to let LinuxCNC know how far behind the drives are? We had previously only tried looping back the actual position and it seemed like the drive and LinuxCNC were fighting for control, so we left it open loop.

Thanks for your patience with this, we really appreciate the help!

I don't think that there is a way to use only the following error from the drive other than for diagnostic purposes (such as using Halscope.)
How were you trying to use the position feedback from the drive?
If you only connect it to the "axis.N.motor-pos-fb" (or "joint.N.motor-pos-fb" in master) Linuxcnc will only use it for DRO and calculating joint following error alarms. It should not cause any fighting, provided the position command and position feedback are scaled the same and following errors are reasonable. For there to be any fighting think you would have to add a PID loop.