Positional error of TCP on CB3 UR5

I’m currently having an issue with the reported position of the TCP of a UR5. The robot is reporting several millimeters off from real world measurements (over 15mm in some cases). I’ve tried re-calibrating the joints, but this didn’t seem to do anything.

To test this I’ll place the tool on the work surface and orient it straight up. The z-position will read ~0. If I flip the tool tip so it now faces directly up and measure the distance from the work surface, the error between reported and measured is over 15mm.

Is this normal? Is there something I can do about it?

Have you verified there is nothing mechanically loose with the robot? It’s not normal to be off that far. We have done studies with a UR10 using a dial indicator and normally have repeatability below the 0.1mm that UR reports in the sprlecifications.

We did have an arm though that would occasionally be off by 3-5mm, it had a mechanical issue with a joint causing the robot to be off.

@heidt Do you have an E-Series or CB3 UR5? Which software version is it running?

It’s a CB3 running version 3.3 I believe though I’ll have to double check that.

EDIT: actually it’s running 3.4.5.100 my mistake

As far as I can tell everything seems secure. Is there anything you suggest I check? Thanks!

How exactly did you measure and command the flip? Is zero reference also Z=0 for the robot, e.g. the reference plane is the robot mounting plane? Can you send a picture of the tool in flipped position? Do you see errors in X,Y also? Or just Z?