Placement Accuracy

geo0rpo

unread,

Jan 9, 2022, 12:43:43 PM1/9/22

to OpenPnP

Hello everybody!

Please look at the pictures below.

You can see the the pipeline recognizes the qfp part properly, the machine know where to place it but still I have this error.

I would say that this is not a matter of backlash because no matter how many times I jog the head to all directions at max speed, the head always returns spot on the correct position of the image above.

Of course I have double checked camera-nozzle offset and also tip runout is supposed to be OK. Does anyone have any tips or ideas?

geo0rpo

unread,

Jan 9, 2022, 12:47:51 PM1/9/22

to OpenPnP

Bottom camera is also calibrated from the tips menu:

and also units per pixel and lens calibration are also OK.

mark maker

unread,

Jan 9, 2022, 1:03:34 PM1/9/22

to ope...@googlegroups.com

and the bottom camera focal plane is at PCB surface height?

_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/096e91f5-8c8e-4a67-ac97-493e0ee3881an%40googlegroups.com.

geo0rpo

unread,

Jan 9, 2022, 1:04:24 PM1/9/22

to OpenPnP

And also the error is repeatable.

I tried many times and I captured an image each time. The error is exactly the same.

geo0rpo

unread,

Jan 9, 2022, 1:06:30 PM1/9/22

to OpenPnP

>and the bottom camera focal plane is at PCB surface height

If I understand the answer correctly, then NO. The bottom camera focal plane is at Z0 which is about 15mm above pcb height.

Jarosław Karwik

unread,

Jan 9, 2022, 1:07:31 PM1/9/22

to OpenPnP

So small tip for the chip ? Maybe the chip slips during the final placement move ?

geo0rpo

unread,

Jan 9, 2022, 1:12:08 PM1/9/22

to OpenPnP

Yes I thought someone would ask.

The picture with the 503 tip is just to show the bottom camera calibration.

Of course I use a 506 tip.

mark maker

unread,

Jan 9, 2022, 1:13:08 PM1/9/22

to ope...@googlegroups.com

But you did the nozzle-to-camera offsets at PCB surface level?

Using Issues & Solutions?

https://youtu.be/md68n_J7uto?t=462

https://github.com/openpnp/openpnp/wiki/Calibration-Solutions#calibrating-precision-camera-to-nozzle-offsets

_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/7a1a229e-0846-402a-84d2-a88f5103085dn%40googlegroups.com.

geo0rpo

unread,

Jan 9, 2022, 1:27:33 PM1/9/22

to OpenPnP

I do not have this option.

I have the 550c version.

geo0rpo

unread,

Jan 9, 2022, 1:32:54 PM1/9/22

to OpenPnP

Something that puzzles me is that the bottom camera view look rotated but the rotation value is 0.

mark maker

unread,

Jan 9, 2022, 1:33:10 PM1/9/22

to ope...@googlegroups.com

Note: if you leave "Fundamental" solutions open, you will not see the subsequent solutions, that are dependent.

Please work them through, in the order they are suggested. If you have solved some, press Find Issues & Solutions again (and again), to get the dependent ones. I cannot give better advice than what Issues & Solutions can.

Watch the whole video perhaps:

https://youtu.be/md68n_J7uto

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/4549348b-6f44-4629-9bbd-1ca36d80c812n%40googlegroups.com.

mark maker

unread,

Jan 9, 2022, 1:34:34 PM1/9/22

to ope...@googlegroups.com

> Something that puzzles me is that the bottom camera view look rotated but the rotation value is 0.

All this is solved, if you go through Issues & Solutions faithfully.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/4ddb28c8-efce-4627-98de-043629eafd97n%40googlegroups.com.

geo0rpo

unread,

Jan 9, 2022, 1:44:18 PM1/9/22

to OpenPnP

I have tried "some" of the new features but I think that when I did the top camera advanced calibration then I had a small error recognizing the pcb fiducials

so I disabled the advanced calibration.

Anyway I will try all issues and solutions again.

Is there a "proper" sequence for them?

mark maker

unread,

Jan 9, 2022, 1:46:07 PM1/9/22

to ope...@googlegroups.com

> Is there a "proper" sequence for them?

Top to bottom.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/e832fa16-9572-4f4a-bcd3-80cc23c3cc16n%40googlegroups.com.

mark maker

unread,

Jan 9, 2022, 1:51:25 PM1/9/22

to ope...@googlegroups.com

> I did the top camera advanced calibration then I had a small error recognizing the pcb fiducials

That is expected, if the primary calibration fiducial issue is still open.

The fact that advanced calibration is even suggested, means that you must have manually set the calibration fiducials on the head. While this is a possible/legal way to do it, the responsibility of doing it right is then on you. 😉

Therefore, I recommend letting Issues & Solutions guide you, so all this is properly sequenced and validated.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/87f938ea-364d-0766-b701-252803d62edd%40makr.zone.

mark maker

unread,

Jan 9, 2022, 1:52:22 PM1/9/22

to ope...@googlegroups.com

I'll be offline now. Sorry.

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/7ef8f869-a6a6-d1b6-47aa-6845c0845897%40makr.zone.

geo0rpo

unread,

Jan 9, 2022, 3:14:05 PM1/9/22

to OpenPnP

>and the bottom camera focal plane is at PCB surface height?

I used to have it this way. I changed it so that the nozzle with the part moves over the bottom camera and just be on focus

thus not wasting time to lower to part down to focus and then go back to Z0. It is much faster now.

I am running all the open issues and solutions, I am about to end and saw a suggestion that the bottom camera focal plane should be to pcb level which is -15mm.

Please tell me how important this is because the machine is much faster now.

For now, I continued the issues and solutions with my current focal plane at Z0.

...On a second thought this compensates for the nozzle shaft not being absolutely vertical?

Thanks.

geo0rpo

unread,

Jan 9, 2022, 3:26:54 PM1/9/22

to OpenPnP

I also noticed that the advanced calibration has a method for detecting the tip center that is much more precise than the tip pipeline.

Just saying...

geo0rpo

unread,

Jan 9, 2022, 3:36:14 PM1/9/22

to OpenPnP

tonyl...@gmail.com

unread,

Jan 9, 2022, 3:43:17 PM1/9/22

to OpenPnP

Both camera and Z-axis tilt (measured relative to the machine X-Y plane) contribute to horizontal offset errors. Advanced camera calibration attempts to correct for the tilt of the cameras but we still need a calibration to correct for Z-axis tilt. Until we have that, it's best to keep everything as close as possible to the same Z height.

mark maker

unread,

Jan 10, 2022, 3:41:32 AM1/10/22

to ope...@googlegroups.com

> I used to have it this way. I changed it so that the nozzle with the part moves over the bottom camera and just be on focus thus not wasting time to lower to part down to focus and then go back to Z0. It is much faster now.

Yes, that's a valid argument. But...

> ...On a second thought this compensates for the nozzle shaft not being absolutely vertical?

Yes. Having the focal plane at the Same Z as the PCB makes sure there is no error brought in by a nozzle axis not being perfectly parallel to Z.

However, the Nozzle Tip Calibration aims to compensate for any such error, even if multiple nozzles are not perfectly parallel among themselves:

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

https://github.com/openpnp/openpnp/wiki/Nozzle-Tip-Calibration-Setup

So in theory it is valid to have a different Z level for the focal plane. But I guess in practice every mechanical system still has slightly larger errors that way. Even professional systems use the "same Z plane" principle it seems, so I guess there are very good reasons:

https://youtu.be/tJ-N-6DMIe4

However, I'm still confident, once you finished your calibration "the Issues & Solutions way", the errors will mostly be gone. Make sure to enable Pre-rotate!

Note: there have been on-and-off reports about the PCB fiducial calibration sometimes acting unexpectedly. I don't know much about that side of the equation, @Tonyluken would be the expert there.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/8f9db384-1685-4993-a34e-53b178284778n%40googlegroups.com.

mark maker

unread,

Jan 10, 2022, 3:50:05 AM1/10/22

to ope...@googlegroups.com

> I also noticed that the advanced calibration has a method for detecting the tip center that is much more precise than the tip pipeline. Just saying...

Yes, if you use the Issues & Solutions calibration, you will automatically get that pipeline assigned to your calibration tip.

The DetectCircularSymmetry stage is a no-tuning probabilistic stage of mine. It does not need any user input except the well-known physical diameter of the feature you want to detect. No thresholds etc. needed. Btw. the same pipeline is also used for the fiducials and the "confetti". It seems to work very well so far, if users keep confirming it works, I will make it the new default in OpenPnP for nozzle tips and fiducials.

mark maker

unread,

Jan 10, 2022, 3:50:54 AM1/10/22

to ope...@googlegroups.com

No bad, right?

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/df564cd7-f7ec-49b9-87a8-7610fccf5917n%40googlegroups.com.

jdlv

unread,

Jan 10, 2022, 4:09:52 AM1/10/22

to ope...@googlegroups.com

Hi,

had unexpected placement accuracy troubles too, now solved with the new 3D camera calibration.

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/44612d66-ab07-37b2-4346-6568c7a080a7%40makr.zone.

geo0rpo

unread,

Jan 10, 2022, 5:09:02 AM1/10/22

to OpenPnP

>No bad, right?

Are you kidding? It is perfect!!! Well done!!!

>Yes, if you use the Issues & Solutions calibration, you will automatically get that pipeline assigned to your calibration tip

I finished all issues and solutions and pressed again (and again) the find issues but all issues are now solved. I think that the tip pipeline has not changed though.

I will check and if not I will copy-paste it as you explain.

It also worked perfectly for the confetti whenever the confetti decided to stay down, since I do not have part blow off. A little water underneath and longer place time helps.

But yes, this new DetectCircularSymmetry of yours is very good!!! But how can I measure the tip vision diameter? And is it in pixels or mm?

Mark, btw I tried to reach you by email. Please just confirm if you received.

Thanks.

geo0rpo

unread,

Jan 10, 2022, 1:02:51 PM1/10/22

to OpenPnP

Hello friends!

Well, I replaced all of my tips pipeline with Mark's DetectCircularSymmetry and it works very good.

What is a good "score" to aim at? I am trying trial and error in the vision diameter because I cannot measure the inner hole of the tip.

mark maker

unread,

Jan 10, 2022, 2:27:45 PM1/10/22

to ope...@googlegroups.com

> I cannot measure the inner hole of the tip.

Yes you can, I've added it to the Wiki (including a screen-recording-gif):

https://github.com/openpnp/openpnp/wiki/DetectCircularSymmetry#nozzle-tip-calibration

Measuring the Vision Diameter using the camera and relative DRO mode:

Center the nozzle tip.
Zoom the Camera View using scroll-mouse.
Jog to the right edge of the wanted circular feature.
Click the DRO to put it in Relative Mode (it turns from greenish to blueish).
Jog to the left edge of the wanted circular feature.
Read the diameter off the DRO X coordinate.
Enter as Vision Diameter.

Watch the gif:

Measuring the nozzle tip feature diameter avi

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/1a6aed31-8373-437e-a485-3c202d263baen%40googlegroups.com.

tonyl...@gmail.com

unread,

Jan 10, 2022, 2:49:13 PM1/10/22

to OpenPnP

And note that you want the measure the outer diameter of the portion of the nozzle tip that contacts the part during a pick. The inner hole can be off center relative to that, but it doesn't matter - what you care about is how eccentric the motion of the portion of the nozzle tip that is actually moving the part.

geo0rpo

unread,

Jan 10, 2022, 3:08:52 PM1/10/22

to ope...@googlegroups.com

understood. thank you guys!

Also I enabled pre rotate in bottom vision settings and I noticed that when I tried to test alignment of a part, it first rotated to various angles before even trying the pipeline. I do not know if this is correct and I must say that it worked OK even before enabling pre rotate. it was in default.

thanks again

You received this message because you are subscribed to a topic in the Google Groups "OpenPnP" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/openpnp/AocvRt3Qz7o/unsubscribe.
To unsubscribe from this group and all its topics, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/abdb5666-9615-4897-a051-ab08a5e19d55n%40googlegroups.com.

mark maker

unread,

Jan 10, 2022, 3:08:55 PM1/10/22

to ope...@googlegroups.com

> you want the measure the outer diameter of the portion of the nozzle tip that contacts the part during a pick. The inner hole can be off center relative to that...

I would argue that the outer material can well hang over the edge of the part, but the inner hole must never do that, or it will leak. So I would argue that in case the inner hole really is significantly off center, and parts are so damn small that it matters, the inner hole is actually the better reference. 😁

Exaggerated illustration:

The same is true, if the front if the tip is not flat.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/abdb5666-9615-4897-a051-ab08a5e19d55n%40googlegroups.com.

mark maker

unread,

Jan 10, 2022, 3:10:26 PM1/10/22

to ope...@googlegroups.com

> it first rotated to various angles before even trying the pipeline

It should definitely not do that. Are you sure?

Log?

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/CANKvoS0uxMCx5Uq%2B_W%3DPBr6uhqh7An9kRMwrSBMkZrRAudFAbg%40mail.gmail.com.