Bottom Camera advanced calibration not working :-(

314 views
Skip to first unread message

Fabian Kropp

unread,
Feb 8, 2023, 1:01:49 AM2/8/23
to OpenPnP
Hi,

i had some hope, that settling time may also case an issue on the bottom camera for advanced calibration. 
It looks like the nozzle cannot be found.
See attached logs / config / screenshot. (beware the screenshot is older- forgot to make a new one)

When I go to nozzle tip calibration and check the pipeline it detects the nozzle tip without issues. 

Any clue where to start or where to debug any further?

BR

Fabian

P.S.: the machine is a LumenPNP V2 from opulo

 
OpenPnPBottonCalibrationFidNotFound.log
Advanced bottom camera fail.jpg
vision-settings.xml
packages.xml
parts.xml
machine.xml

tonyl...@gmail.com

unread,
Feb 8, 2023, 4:58:32 PM2/8/23
to OpenPnP
First, I have to ask - what is the largest sized part you want to image with your bottom camera?  The reason I ask is that your field-of-view appears to be very small (assuming that nozzle tip is about 1 mm in diameter, it looks like you only have about 4 mm by 2 mm of viewable area).  If you want to handle larger parts, I suggest you fix that first by either changing to a shorter camera lens or lowering the camera.

If your intent is to only handle very small parts and want to keep that small FOV, I think you may have uncovered a short coming of the Advanced Calibration algorithm.  It basically attempts to measure the units-per-pixel by moving the nozzle tip +/-0.5 mm in a zig-zag pattern over the center part of the image and expects to find the nozzle tip still pretty close to the center of the image. In your case, since the image is so zoomed in, that 0.5 mm of motion is moving the nozzle tip too far from the image's center to be detected. One thing you could try, is quit out of OpenPnP and manually edit your machine.xml file with a text editor.  Find the Bottom camera section - look for a line that starts with:

<camera class="org.openpnp.machine.reference.camera.OpenPnpCaptureCamera" id="C2" name="Bottom" looking="Up"

and below that, that find the line that looks like this:

<trial-step value="0.5" units="Millimeters"/>

Edit the 0.5 and change it to maybe 0.1 and save the file. Restart OpenPnP and retry Advanced Calibration on your bottom camera again.  It may or may not work but it's worth a shot.  It may be that the nozzle tip is just too large for that camera's FOV.

Tony

Fabian Kropp

unread,
Feb 8, 2023, 5:18:21 PM2/8/23
to OpenPnP
Hi Tony, 

thank you for the analysis.

regarding the camera position: it's a flaw of the Lumen PNP V2 I think.

Lowering the camera will be a step I would take on later. I need to if I want to convert to dual head.
I'll give your suggestion a shot tomorrow and report back here for anybody else who is struggling with a lumenPNP


BR

Fabian 

mark maker

unread,
Feb 9, 2023, 3:04:05 AM2/9/23
to ope...@googlegroups.com

> regarding the camera position: it's a flaw of the Lumen PNP V2 I think.

I wish those guys would talk to the actual OpenPnP community, i.e., including some of us active developers, before they multiply their stuff. It is IMHO not asked too much for them to thoroughly go through I&S once in a while for their machine design. So issues like this would fly in their faces early.

And they could then also report to us if OpenPnP / I&S does not properly address their machine's capabilities. It could only be win-win!

_Mark

--
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 on the web visit https://groups.google.com/d/msgid/openpnp/7addaf8a-35a6-4318-aff6-9ab000360559n%40googlegroups.com.

tonyl...@gmail.com

unread,
Feb 9, 2023, 9:57:53 AM2/9/23
to OpenPnP
What kind of lens does the camera have?  Maybe you can easily change it out to give you a larger FOV.

Tony

fabian kropp

unread,
Feb 9, 2023, 10:01:59 AM2/9/23
to ope...@googlegroups.com
Unsure, but lowering it by 30 mm which is the mount for the lumenpnp v3.

That was on my list anyways for the future.
But i would like to postpone this for a bit as I need to get some pcb finished soon.

tonyl...@gmail.com

unread,
Feb 9, 2023, 11:28:48 AM2/9/23
to OpenPnP
How far is the camera from the nozzle tip now? For instance, if it is currently 30 mm from the nozzle tip, lowering it another 30 would double your FOV (in each direction). If it is currently 60 mm from the nozzle tip, you would only gain 1.5 times your current FOV. On the other hand, if it is currently 10 mm from the nozzle tip, you would quadruple your FOV.

Fabian Kropp

unread,
Feb 9, 2023, 11:29:55 AM2/9/23
to ope...@googlegroups.com
More like 5 mm :-) haven’t measured though



Fabian Kropp

unread,
Feb 9, 2023, 11:35:30 AM2/9/23
to ope...@googlegroups.com
Currently pondering of creating a poor mans solution to space down the bottom camera …. 

tonyl...@gmail.com

unread,
Feb 9, 2023, 11:43:01 AM2/9/23
to OpenPnP
> More like 5 mm

Ahh, then lowering by 30 mm should expand your FOV by about 7X (to about 28 x 14 mm) - assuming you can adjust the focus so that the nozzle tip remains in focus at its current position. You might be able to gain some more by adjusting the focus point even higher (and redoing the auto focus stuff).

fabian kropp

unread,
Feb 9, 2023, 12:11:51 PM2/9/23
to ope...@googlegroups.com
The issues and solutions told me to put the focus point to pcb level that’s why it is so close.
If that is not required I just disassembled my machine for nothing 

tonyl...@gmail.com

unread,
Feb 9, 2023, 12:26:12 PM2/9/23
to OpenPnP
I don't think having it at PCB level is absolutely necessary - Mark could answer better though.

mark maker

unread,
Feb 9, 2023, 12:50:39 PM2/9/23
to ope...@googlegroups.com

> The issues and solutions told me to put the focus point to pcb level that’s why it is so close.

It's not an absolute necessity, just a recommendation for a machine design that is "naturally" cancelling out the errors. The alignment will take place at the same Z as the placement, so any inaccuracies in the nozzle Z axes will be irrelevant.

The extended OpenPnP Nozzle Tip Calibration can compensate a nozzle specific Z axis tilt/offset, so it is probably not that important anymore:

https://makr.zone/improved-runout-compensation-and-bottom-camera-calibration/346/

I don't know how effective and realistic the compensation is, but some users have reported it works well.

_Mark

Fabian Kropp

unread,
Feb 9, 2023, 1:13:39 PM2/9/23
to OpenPnP
i startet printing the parts, as I had severe placement issues, I would try to be at least in close proximity to the PCB level (which is a tad higher than the intended height for the lumenPnP (1.4 mm)

With the adjusted test steps it at least finished the calibration for now. 
But I will restart the calibration from scratch anyways as soon as I finished my production runs.

Thanks for effort

Fabian 

fabian kropp

unread,
Feb 10, 2023, 2:31:52 AM2/10/23
to ope...@googlegroups.com
Thanks for all the help.

Moved the camera down now.
I am curious what breaks next 🤣

BR

Fabian 

Reply all
Reply to author
Forward
0 new messages