Is automatic re-calibration of the machine possible using multiple homing fiducials??

[Greg Toman]

Is it possible to use more than one homing fiducial and apply the same corrections to the entire machine, as is done with the fiducials on a pc board?

My machine uses belts, and occasionally I have had to re-calibrate the machine. It finds the single homing fiducial just fine, but probably due to the belts stretching over time, I occasionally have to  reset the pick location for most of my feeders.  This is time consuming. Is there already a way in OpenPnP to do this?

It seems like others who use belts must have encountered this same problem. I can see this drift  easily as I placed a second fiducial at the location where the head is parked. Over time the accuracy of the park location drifts, which is also reflected in the accuracy of pick locations from the feeders.

Thanks
Greg

[Greg Toman]

Interesting that there was no response to this. Since I posted this, I tightened my belts (my machine is about 700mm x 700mm and I use coreXY), and the drifting at the park location after homing seems to have diminished significantly to the point where I can't really measure any. 

Originally I had used a method to set the belt tension according to what was the limit according to the side loads on the stepper motor. This was about half of the belt tension recommended for the belts (GT2 belts).   I messed some things up, and and to re-adjust the belt tension, so I just re-tensioned the belts by feel (probably a little tighter than it was), and the drifting seems to be gone.

So maybe this is useful to others. If you get drifting of the park position after homing from week to week, this might be a sign you need to tighten the belts a bit more. 

Great software. I hope this helps someone.

Greg

[Toby Dickenson]

Hi,

Sorry I missed this first time around.

It is pretty normal that the absolute positioning of the machine might drift over time. But this drift is expected to be small. The absolute pick position of a feeder will drift, but it should be far less than the offset that would affect the viability of the pick.

Positions that need precision are calibrated for each job. This is one of the reasons why it is recommended to have lots of fiducials on your board. Not because we expect the board to be out of square, but because the machine axes might be.

Belt tightening makes sense.

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openpnp/c513053a-0995-4ffa-83ae-8c4dd5e22da0n%40googlegroups.com.

[Greg Toman]

So after a couple of weeks, and probably some change in the temperatures of my shop, i did notice a slight bit of drift in my machine again. That is, it homes visually to very high accuracy, but when it parks (about 400mm away in Y, 200mm in X from the home location), I see it the park position vary by up to 0.2mm  in each axis. You are correct that this is tolerable for picking the part from the feeder, but if it was to drift by about double this amount, it would be a problem. 

So I guess I am back to my original impression that it would be beneficial to have the machine calibrate itself using multiple homing fiducials, I can't speak to how much effort this would take, but I know I would like it!

I downloaded the code, and while I have written a little software in java, I have to admit it is a bit overwhelming for a non-software guy like me.

I have to say that I am very impressed with the placement accuracy that can be achieved using 4 fiducials on a pcb. I always design all my boards to have 4 fiducial now, and consider it mandatory for boards that have fine pitch parts. The accuracy of this feature is what inspired my thoughts to include this for the entire machine also.

Thanks for great software

Greg

[Artem Stanchak]

I have a regular XY kinematics. I also have belts and everything is fine, there is no offset. Previously, I used a nema17 and GT2-6 mm belt. Then I improved the design to nema23 and 10 mm belts. So try to review your mechanics. P.S. previously there was a backlash on nema17, namely when the motor was on the brake. And also try to calibrate the axes.

неділя, 14 грудня 2025 р. о 21:17:25 UTC+2 n26...@gmail.com пише:

[bert shivaan]

Back in R1, there was a guy here that needed to re home his machine due to temperature creep. So he wrote a script that ran at the beginning of every job to home X and Y visually. It would run the machine to has humming dot I think and proceed to home. It was fast because he did not need to wait for homing speeds to find the dot.

I am not sure if this is helpful or what you need.

I agree OpenPNP is a huge and overwhelming set of software.

To view this discussion visit https://groups.google.com/d/msgid/openpnp/f6e72332-f599-48f9-bfa2-02940c0a27d5n%40googlegroups.com.