Secondary Z fiducial nozzle misalignment

[Litterio Andrea Guainella]

Hi guys,

I am using latest testing openpnp version so I renamed .openpnp2 folder to .openpnp2.bck for cleaning installation.

In coincidence to this I printed new stripfeeder as mention here with little custom configuration so Z pcb is different than Z stripfeeder about 22mm.

After doing all steps from "Issues & Solutions" I see offset nozzle about 2mm in Z stripfeeder as image.

I suppose that mechanical frame is correct aligned (as tested withbubble level) in Z stroke also in movement in all position (front/back and left/right).

I know when possible place all to Z primary fiducial (pcb, strip feeder, automatic feeder, etc...) but in this case it would be great to have pcb and strip feeder at different Z level.

Also happen with advanced camera calibration or not.

I searched in group if same issue exsist and the most similar is refer to this post but I have checked that in machine.xml " disable-tangential-distortion-correction="true" " is 

properly configured.

Any suggest is appreciate. Thanks.

LAG

[mark maker]

Hi LAG,

I read your post a few times, but I do not understand what the actual problem is?

The images look good, the virtual Z is indicated on the camera view as -20.1mm, which means your PCB would be at ~-42mm which sounds about right for a Liteplacer, that I guess this is. The ticks on the cross-hairs look good, they seem to align 4mm with the sprocket hole pitch.

So everything seems to work alright. What is the problem?

Obviously the image is a bit blurred but that should not bother computer vision. If not, try this alternative pipeline here, assuming you use the ReferenceStripFeeder:

https://github.com/openpnp/openpnp/wiki/DetectCircularSymmetry#referencestripfeeder

_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/db4473e6-0704-4951-ad14-f21b02d654ecn%40googlegroups.com.

[Litterio Andrea Guainella]

Yes Mark you are right I have a Liteplacer with TinyG sorry for do not  mentioned before.
As you can see from 2nd image the nozzle not centered in part when I simulate pickup.

The horizonal black band was caused from led bar near table (the ceiling is lower than normal and this affects the camera vision).

Now I haven't problem with cvpipeline so I just used default but thank you for your suggest.

The problem is nozzle misalignment at Z stripfeeder from down camera position correct in cross-hairs.

I try to put strip-parts at Z pcb level and nozzle alignment is correct for pickup action.

The current configuration primary Z fiducial (rig calibration) is at Z pcb level and secondary Z fiducial (rig calibration) is at Z strip feeder level.

The vision cvpipeline seems work properly with detect strip-parts but when click to pickup this part nozzle have offset about +1,3mm in X as you can see in previous post.

Seems an units for pixel problem but I don't know as solve

LAG

[mark maker]

Ah, now I understand. So the problem might either be a tilt in the Nozzle Z axis or the camera Z axis (view axis).

But tilt in the camera is unlikely, as it happens with both Advanced Camera Calibration and without. Please confirm that you performed the camera calibration with the newest version, i.e. after the disable-tangential-distortion-correction="true"was established. Right? (it is not sufficient to update, you need to perform the calibration again, after updating OpenPnP).

Then just to be sure, let's confirm it is not just a problem of the feeder's specific programming:

Center the camera on the feeder (like the corner of a pocket or something you can judge), then use this button:

Does it show the same offset?

If yes, then tilt in your Nozzle is almost sure.

I don't think you can adequately measure this with a bubble level. In fact I don't know an easy way.

@Others, how do you measure if the Z axis is not tilted?

20mm difference in Z is a lot. I can only repeat: try to use the same Z as for the PCB.

Why do your strip holder need to be 20mm high? This takes a lot of 3D printing time!

Have a look at BlindsFeeder for contrast. They're super fast to print, add minimal Z height (only as much as the part height) and can (optionally) cover the parts, so they don't fall out, are faster to pick (vision only once per job), can optional have OCR/Barcode/QR code.

https://github.com/openpnp/openpnp/wiki/BlindsFeeder

And despite the Liteplacer being ready, you can't use Contact Probing, because the tape strips are only suspended on the sides, and would be poked-through if probed, or parts would surely topple out from vibrations.

https://github.com/openpnp/openpnp/wiki/Contact-Probing-Nozzle

I know, you already printed the feeders... but maybe you should still reconsider 😉

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/b87cb17d-d80b-4c9d-9f28-8dc8ba85c6a2n%40googlegroups.com.

[Litterio Andrea Guainella]

> (it is not sufficient to update, you need to perform the calibration again, after updating OpenPnP).

Yes I know and as mentioned before I renamed .openpnp2 folder (after install new testing version) for cleaning installation so I repeated all steps from "Issue & Solution" avoid any problem.

>  Center the camera on the feeder (like the corner of a pocket or something you can judge), then use this button:

I try also this test but the offset is same both with:

and with:

as you mention.

As I am writing this message openpnp show popup with update is available so I procedeed and now I have this version:

should be last version. For avoid any problem I deleted .openpnp2 folder before installation.

> I know, you already printed the feeders... but maybe you should still reconsider 

NO problem I don't have to use stripfeeder printed (only 2 ganged for test) so I have already started to print blindfeeder (I hadn't really thought about it).

In the meantime I configure openpnp which is very long procedure especially for advanced calibration top, bottom and backslash.

I'll update you as soon as it's done. Thank you so much for suggest.

LAG

[tonyl...@gmail.com]

Wow, a 2mm horizontal offset over 22mm of vertical motion would mean you have more than 5 degrees of tilt!  That seems unlikely to me.  Are you sure you have the tape width configured correctly?  If you have it set to 12mm instead of 8mm, that would explain the extra 2mm offset to the side.

> I don't think you can adequately measure this with a bubble level. In fact I don't know an easy way.

If we did the confetti trick at both the primary and secondary fiducial height we could measure it😉

Tony

[mark maker]

tony> If we did the confetti trick at both the primary and secondary fiducial height we could measure it😉

After Advanced Camera Calibration, and assuming it did detect the camera view axis tilt accurately, yes! It would be easy to add. I'll keep that in mind.

LAG> For avoid any problem I deleted .openpnp2 folder before installation.

You really do not need to do that each time, unless you completely wrecked the config. Usually, just reopening all the subsequent Issues & Solutions items is sufficient.

For most things, Issues & Solutions is permanently monitoring your settings, i.e. the suggestions are not "one-shot". If settings do not match according to its logic, it will alert you, again. This is true for maybe 80% of the settings it can propose.

For the rest, there are some (like the calibration steps) that need to be re-opened explicitly, if the machine has changed, because it cannot detect that. Very few solutions are "one-shot": Issues & Solutions will only propose them if they are empty/missing. This is for machine specific stuff, where I&S can only guess. Like some machine specific G-code commands. If in doubt just delete these and a new proposal will come.

So it is only these few that you need to keep in mind.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/f159d11f-88e5-48fe-9da4-0290172c862dn%40googlegroups.com.

[tonyl...@gmail.com]

> After Advanced Camera Calibration, and assuming it did detect the camera view axis tilt accurately, yes!

Actually, it could be done before Advanced Camera Calibration as well assuming you add the Z-axis tilt compensation (and change the X/Y backlash compensation to be applied before negative Z motion and after positive Z motion).  That should work even with a tilted camera axis - but you would still need to redo the nozzle offset calibration after Advanced Calibration.  Even if there is any remaining camera tilt after Advanced Calibration, it would simply be compensating for the difference in tilt between the camera and the Z-axis.

Tony 

[mark maker]

I don't understand. It is true that the down-looking camera tilt and the nozzle tilt cannot be distinguished from two Z level confetti calibrations, right?

For example, these two (exaggerated) constellations will exhibit the same apparent delta:

Or do you suggest to first just determine the overall tilt and then only adjust (as a computation) the nozzle tilt for the camera tilt, once the Advanced Calibration was done? That would be possible, I guess.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/213325b1-dd48-4b7d-8a04-f5d82b08f88cn%40googlegroups.com.

[bert shivaan]

At the very high risk of not knowing fully what the issue is here, I have a question.

Could the nozzle tilt not be found by simply parking the nozzle over the bottom camera, take a snap shot at z fully retracted, then extend z fully and take another snap shot. This would then show how much Z is not perpendicular to the camera.

Would that at least answer 1 question?

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/fec660c3-8602-cbb2-a1ce-1a671d82471c%40makr.zone.

[tonyl...@gmail.com]

cncmachineguy - Yes, the bottom camera could be used to determine the Z-axis tilt - in fact Advanced Calibration of the bottom camera already logs the necessary information to determine the tilt of the Z-axis. But it's not a very direct measurement and is not really measuring what we really care about - that is the top camera-to-nozzle offset.

Mark - I think you illustrated it perfectly - it doesn't really matter whether one or the other or both are tilted - in terms of camera-to-nozzle offset, they are indistinguishable.  But we really don't care - all we really care about is the offset between the nozzle and the camera.  Once you capture that offset at two different heights, you can compute the offset at any height.

Tony

[bert shivaan]

But if we want to use the nozzle to open the blinds, do we not need to know where it will hit when extended?

Or are you saying we will know that because no matter if the camera is off or the Z, or both, we can then hit the correct spot because we looked at it?

Would it fail if the spot we are trying to hit is calculated position instead of captured by vision?

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/e33ce1c8-a8b2-4361-9170-3b272df21ba8n%40googlegroups.com.

[mark maker]

@Tony> it doesn't really matter whether one or the other or both are tilted

Yes this is true (but see cncmachineguy's valid point, later).

You kind-of make the camera view axis (with whatever tilt it has) the "reference", all the machine's nozzles have to adhere to it.

But you have to adjust all these nozzle specific tilts, once you perform the Advanced Camera Calibration, as it's tilt compensation selects another "pixel ray" as the camera view axis. I.e. you get this:

Or are you saying we should abolish its tilt compensation?

@cncmachineguy> Could the nozzle tilt not be found by simply parking the nozzle over the bottom camera, take a snap shot at z fully retracted, then extend z fully and take another snap shot.

Same problem. Camera tilt is not distinguishable from nozzle tilt.

But admittedly, it would be easier to implement (compared to the confetti trick).

@cncmachineguy> Would it fail if the spot we are trying to hit is calculated position instead of captured by vision?

You have a valid point there. Whenever OpenPnP would make a move based on taking X, Y from another Z and then perpendicularly extending from this to another wanted location, this will incur the camera tilt error.

In case of a (tall) BlindsFeeder, where it takes the fiducial X/Y from down at the table, but then wants to push the cover edge up at part height, based on "blind" geometrical knowledge of how the feeder is made, this will falsely add the camera tilt. In case of the BlindsFeeder this is not a problem, because it knows it has to cope with deft errors from 3D printing and adds a 2mm tolerance when approaching, and gives you edge calibration with a vision feed-back loop, i.e. it does not do it "blindly" in the end.

But for other truly "blind" geometric moves with tall structures, this might become a problem. I'm thinking about nozzle tip changers etc.. So I still think it would be better to simply first calibrate camera tilt using Advance Camera Calibration (that I now find trustworthy after the "tangential" fix) , and then nozzle tilt to that reference. So we get true motion geometry (or as "true" as can be achieved).

Backlash...

Note, as @Tony already mentioned, this needs special treatment for Backlash Compensation in X/Y. It effectively has to take the final location into account when calculating the X/Y move at Safe Z, i.e. it need to already include the nozzle tilt compensation up there, so any backlash counter-move can be done "up in the air" where it is harmless (not down when pressing a part into solder paste). Afterwards, only the Z axis will go down, no tilt compensation along the way.

Folding the backlash counter-move into the downwards-move (saving the extra time) would be the "luxury" solution. But because these two concerns are currently (very) separated, it will probably not be done anytime soon.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/CA%2BKNHNxwynuS2Z%2BqV0wH%3DG7XBdmehw8gqTjyHG56vJ9mieki5w%40mail.gmail.com.

[tonyl...@gmail.com]

Don't get me wrong - we would eventually still need to calibrate the camera to get its axis vertical and then calibrate the Z-axis tilt so that both end up being vertical.  I was just trying to say that during the initial coarse calibration (before Advanced Calibration has been run) that the nozzle offset could be measured at two different heights and compensation applied which would effectively make the nozzle Z-axis parallel to the camera's axis.  That can't be any worse than what we currently do during the initial coarse calibration (assume both the camera and the nozzle Z-axis are vertical), right?

Tony

[mark maker]

> That can't be any worse.

True.

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/fa576c5c-f8ae-42cf-be55-ba26b10cb2ben%40googlegroups.com.

[tonyl...@gmail.com]

Sorry, I didn't mean to derail the OP's question. It would be nice to hear back if the problem is fixed or not. 

[Litterio Andrea Guainella]

>  Sorry, I didn't mean to derail the OP's question. It would be nice to hear back if the problem is fixed or not. 

To completely solve the problem it will be necessary to correctly align all three axes in reference to the installation plane.

When I do I will I will know if higher position than level Z of strpifeeder work properly for pick up part.

In the meantime I am following Mark suggest about use blindfeeder but i am having several problems to print it because I've never printed PETG so I am doing some tests and I hope to succeed soon.

I have the feeling of solving with blindfeeder because it has little thickness. As previously said, if I use a tray approximately at Z level all is ok.

LAG

[tonyl...@gmail.com]

Did you double check that you have the tape width set correctly?

[Litterio Andrea Guainella]

Yes. It is correct. I also check it. 

With down camera never problem seems with alignment as you can see in previous posts

LAG

[mark maker]

> i am having several problems to print it because I've never printed PETG

Do you use adhesive spray? And enough of it? Once I did that: all problems gone.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/0add3af6-21a0-4f01-8bee-d41765cb1892n%40googlegroups.com.

[mark maker]

Just to be clear I meant that special 3D printing spray, not "adhesive" spray.

That's the one I'm using (that's just the first one I tried and it worked well, others may too):

https://www.3dlac.com/

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/481849c1-ebe4-f255-c586-63d885f54aed%40makr.zone.

[Litterio Andrea Guainella]

Sure Mark. I know ... I ordered 3dlac because I didn't have it ...
I am trying with Splend'or (hair spray ~ cheapest way) waiting for 3dlac to arrive.
When finish I will post some image for details

LAG

[Litterio Andrea Guainella]

I was not lucky in 3D print BlindFeeder with petg because the result is not good.

I made a lot of test prints (with only one rail for R0603) and many slicer used (cura, simplify3d, prusaslicer) following Mark settings but I'm not satisfied.

Maybe because I don't have a professional printer (I have Voxelab Aquila X2) also I am not expert with 3D print.

In a desperate attempt to resolve with printer I bought and apply this BLTOUCH thinking it was a bed problem and this PEI instead my glass bed.

Many way try and all with bad result in PETG:

  1) bed glass with hard hair spray (no BLTOUCH)

  2) bed glass with 3dlac (no BLTOUCH)

  3) bed glass with hard hair spray (with BLTOUCH)

  4) bed glass with 3dlac (with BLTOUCH)

  5) PEI without BLTOUCH

  6) PEI with BLTOUCH

Thinking it was a print speed problem I tried with (try order): 60mm/s, 50mm/s, 40mm/s, 30mm/s, 25mm/s but more or less always the same result. In short nothing good.

Test #3 (attach 1, 2)

Another test I try to print with PLA same configuration (#3) (attach 4, 5)

It looks a little better but when try to put cover in rail and slide is very friction hard so I think I will abandon BlindFeeder unless some suggestions come up to solve the problems.

Yesterday I saw this video with jazzychad feeder downloadable here and print it with a little bit modification for align to my machine (attach 6).

It seems maybe work. I will try to configure this in openpnp but for the moment I don't know how to do it. I will study for this

Maybe I'll say a nutshell but it's worth saying. To solve the Z alignment problem could a gyroscope / accelerometer like this work?

Has anyone tried this way?

LAG

[Litterio Andrea Guainella]

attach 1

[Litterio Andrea Guainella]

attach 2

Il giorno domenica 12 giugno 2022 alle 15:55:56 UTC+2 Litterio Andrea Guainella ha scritto:

attach 1

[Litterio Andrea Guainella]

attach 4

[Litterio Andrea Guainella]

attach 5

[Litterio Andrea Guainella]

attach 6

[mark maker]

Hi Litterio,

these prints really look scary.

Although I've made quite elaborate stuff (see also the Push-Pull Feeder), I don't consider myself a 3D printing expert, mainly because my Prusa printer just worked (at least once I found the spray is needed).

Sometimes I had stringing when the temperatures were not right, but that looked more like fine "spider webs", not the thick "crazy spaghetti" you get.

Are you sure you got the temperatures right and your printer can do these high temps without an encasing?

That last gray model (different feeder): is that PETG too?

I'm just asking, because PETG is a much tougher, much more elastic and also smoother material than PLA (less friction between moving parts), and according to the link you sent for the other feeder, this one is printed in PETG too, specifically for friction.

So maybe you should still invest more in trying to solve the PETG printing issue.

I assume you know the BlindsFeeder is a completely different use case, than the one you are now printing. BlindsFeeder can only be used for very small runs, where you buy a few dozen of each part. On the other hand, it can use every last one of these parts on even very short cut strips (no waste), which might be important for expensive parts. It is also very compact, you can put many of them on your whole table.

The other feeder is completely different, it only makes sense when you buy whole reels or at least very long cut strips, and you don't mind wasting some of them for the needed cover tape leads etc. Plus it consumes much more space, and works only on the edge of the machine.

_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/1f931b15-56ba-4b82-8129-a4aefd2cab5cn%40googlegroups.com.

[Litterio Andrea Guainella]

Hi Mark,

>  Are you sure you got the temperatures right and your printer can do these high temps without an encasing?

I am sure that my printer (very cheapest) raise to right temperature bed and extruder that I set (tested with thermistor) with a little offset +/- 3°.

I used PETG sunlu 1.75mm with many attempt for test from 190° to 240° (step 10) to extruder and bed from 60° to 80° (step 5).

I don't have a printer case so I don't know what the results will be.

> That last gray model (different feeder): is that PETG too?

No. Printed with sunlu pla+. It is only for test becuse as you said is better PETG than PLA with issue for friction.

Use case for BlindFeeder is clear as I can guess as you said specifically for expensive parts (my case :-) ) and also for reusability without loose part.

I apologize for last post but I didn't mean that the last feeder can replace BlindFeeder but I mean only for print result. 

I will try to print BlindFeeder with a little bit changes for slicer configuration. Maybe my cheapest printer needs tuning for this.

I will post update with some images.

LAG

[Vinay Dand]

Hi LAG,

1) If possible dry out the PETG spool, many time moisture ingresses in the filament and causes this.

2) Can you just put a cardboard carton on top of printer to  crudely form a temporary enclosure for the PETG  material

(But you you need to be alert about any possible misshape - safety issues - while trying a print)

Best Regards,

Vinay

[Litterio Andrea Guainella]

Hi Vinay,

thank you for suggest.

I will try as you said but I don't have an dry machine for filament. I'll have to buy one.

With cardboard carton I did not think about it :-)

LAG

[mark maker]

> I will try as you said but I don't have an dry machine for filament. I'll have to buy one.

Instead of buying ever more things to try and make the printer work, you might consider buying a new printer. 😇

Obviously, I strongly recommend the Prusa, that thing just works!

At €700 (kit) it has its price, I know, but it is a long-term good investment, I already upgraded mine once, you always get the newest model for very little extra money. I bought a second printer and can tell the upgraded one and the new one are exactly the same in all relevant aspects.

@Vinay, as to drying out the PETG, are you sure that can be an explanation for such phenomenons? Wouldn't water rather show up as bubbly surface?(again, I'm no expert).

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/87b2311e-fed7-4824-91fc-df5908216c0fn%40googlegroups.com.

[Vinay Dand]

> @Vinay, as to drying out the PETG, are you sure that can be an explanation for such phenomenons? Wouldn't water rather show up as a bubbly surface?(again, I'm no expert).

Mark, long back, that was one of the things I have tried and it seems to work but I am sure about the science behind it.

Vinay

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/bzLiD6QZIqY/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/8cbc2dc1-3830-91f9-5620-d3003fe01510%40makr.zone.

[Litterio Andrea Guainella]

Hi Mark,

>  you might consider buying a new printer.

You are right but I can't buy before august so I tried with another little changes  😬 😬 (not so expensive)

For major detail I changed bowden (I am using blu CAPRICORN instead default white PTFE).

As @Vinay suggest I tried use cardboard carton (rather uncomfortable). Certainly better but not in a relevant way.

So last change... With nozzle change (ever 0.4mm) everything has magically changed so I obtain this result (IMHO acceptable for test even if it is still stringing present. A little bit)

I'll try to print later 2nd face (this is only cover).

@Mark beyond my feeling what do you think about it?

LAG

[mark maker]

Getting better but I guess not yet good enough.

Are there no 3D printing gurus on the list who could help?

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/e1eb41db-00f8-46de-8adc-7a6edddb6744n%40googlegroups.com.

[Litterio Andrea Guainella]

Yes Mark. Isn't good but acceptable at the moment.

This is 2nd face... This is also acceptable. I just have to clean them up a little because the stringing is still a lot...

With 3d printer I'm hopeless 😅😅

LAG

[Litterio Andrea Guainella]

Hi all,

I resume this thread for update.

I disassemle complete head of liteplacer and reassemble with most accuracy possible.

For this operation I bought an accuracy square (din 875/00) and digital angle sensor for measure position on reassemble parts.

I have only one video because I think was recording instead the phone was powered off by battery loss.

As you can see seems straight when move to -Z and return to +Z 0 position. Also unit for pixel seems correct because all step to move is a 1mm.

Only for remember, the problem is nozzle misalignment at different height of Z placement/pickup position so when camera crosshair point to target then press nozzle button push down to Z target position, the nozzle is not align with camera looked position.

This problem is affect with or without advanced calibration.

As soon as possible I will post new video with two side for detect nozzle straight when move down.

(I will keep the phone motionless for more clarity)

LAG

[bert shivaan]

so the camera is tilted but the z axis moves correctly?

And this tilt can be in the camera itself, does not mean how you mounted it.

--
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/fd944751-3786-410a-8ef7-52d11b15a4een%40googlegroups.com.

[Litterio Andrea Guainella]

I think so too so right now I am changing both camera (up + down) from stock to ELP (ELP-USBFHD01M-L60) but for me is more difficult draw a good holder for 3d printing.
I was try to search in group but I find nothing.

LAG

[bert shivaan]

I think recently Mark and Tony have made code changes to adjust for this.

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/1fa044c3-476e-4605-86ed-1a497d4a5ec6n%40googlegroups.com.

[Litterio Andrea Guainella]

Yes, you are right but as I said the problem is there with or without advanced camera 3d calibration

I hope solve with both camera change

LAG

[mark maker]

Hi LAG,

Try this:

1. I am assuming you re using Issues & Solutions to use the Advanced Camera Calibration. Disable it on the camera and then press Find Issues & Solutions to get it proposed again.
   
2. exit OpenPnP
3. edit machine.xml
4. double-check that you find these settings in both cameras:
   

   disable-tilt-correction="false" disable-distortion-correction="false" disable-tangential-distortion-correction="true" 

5. search for up-looking-secondary-offset-Z-mm=
   
6. change the value (default 2mm) to something larger, where the nozzle tip is still not too blurred and the nozzle tip circular feature still roughly the same size (so it gets detected).
   
7. Maybe a larger Z difference will improve the tilt compensation accuracy. Please report back if possible.
   

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/a55367fd-efd8-4ad6-8aa6-1e0f05c98045n%40googlegroups.com.

[Litterio Andrea Guainella]

Hi Mark,
thank you for your suggest. Now I have just dissassembled both camera so I can't try this procedure.
I will try this when finish mount new ELP cameras.

LAG

[tonyl...@gmail.com]

LAG - How much offset do you have at the pick height and how much different is the pick height than the primary calibration fiducial?  I think earlier in this thread you said it was about 2mm and the difference between your primary fiducial and your pick location was about 22mm vertically.  That works out to more than 5 degrees of tilt (which is a lot of tilt).  Can you also post your TinyG configuration (send the $$ command in the driver console and then exit from OpenPnP and post the latest log file).

Mark - your points 5 through 7 only really apply to up-looking cameras - correct?  If I understand the OP's problem, it is with the down-looking camera and/or the nozzle's Z-axis.

[Litterio Andrea Guainella]

I just finish to mount (with  temporarily support) both camera.

With new camera the misalignment is reduced a lot but is far to goal.

In this configuration Z level and primary fiducial are -18.600, secondary at -9.100 and stripline (for test) at table level (wood base) about -40.000

after advanced calibration with new ELP camera, the top image seems strange (I use 6mm len) instead bottom is normal

I'll try mark's suggestions shortly so I'll post the differences

LAG

[Litterio Andrea Guainella]

Even with the Mark procedure the result is the same.

I changed  up-looking-secondary-offset-Z-mm= from 2.0 to 8.0.

Useless to post the images because they are like the previous post

LAG

[tonyl...@gmail.com]

That's not unexpected - up-looking-secondary-offset-Z-mm is only used for up-looking camera calibration so changing its value shouldn't affect the down-looking camera calibration.

Can you put a spacer under the tape to raise it up to near the primary Z calibration height and see if the offset error improves?  Can you also try it at an even higher height to see if the offset goes the other way?

Tony

[Litterio Andrea Guainella]

Yes Tony. After apply the procedure that Mark said I did up camera calibration advanced.

If I put tape to Z primary fiducial level all is ok also if try to put tape to Z secondary fiducial level.

LAST NEWS: after delete .openpnp2 folder and follow all steps by I&S tadaaa... Now all work correctly. So I can pickup a part in Z primary fiducial level (-18.500), at table level (about -41.000) and at Z secondary fiducial level (-9.000)

I attach machine.xml for investigate but seems nothing has change from last calibration.

I've done a lot of things so knowing what made it work is very difficult

LAG

[Litterio Andrea Guainella]

Just for mention that in recent days I used ever nozzle #1 for any calibration but the last time I used #2 because seems the best for calibration.

Perhaps with nozzle #1 there are many false detection  which would explain the systematic error.

I hope it is useful for liteplacer users.

LAG

[mark maker]

> Mark - your points 5 through 7 only really apply to up-looking cameras - correct?

Correct.

> If I understand the OP's problem, it is with the down-looking camera and/or the nozzle's Z-axis.

Oh, I must have misunderstood. I was aware that the original problem is a difference in nozzle offsets on the two different Z levels. Which might hint at either nozzle Z tilt or (wrongly calibrated) top camera  tilt. But now I assumed he was talking about seeing the nozzle displace in X/Y when looking up with the bottom camera and jogging in Z. After rereading his post, I see that I'm probably mistaken.

So yes, my steps 5 - 7 do apply to the bottom camera and using it to check for nozzle tilt. i.e. seeing whether the nozzle displaces in X/Y when moving in Z.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/7a17be1c-db26-4495-b88e-6c9d73cda476n%40googlegroups.com.

[Litterio Andrea Guainella]

Hi all,

I am sorry to inform you that my enthusiasm died out too soon due to a false alarm.

It seemed that everything was magically solved and instead the problem is precisely the nozzle # 1. 

It's still too early to find the anomaly but I'm working hard to figure out where the catch is. 

I will post further updates in the next few days. :-'(

LAG

[tonyl...@gmail.com]

I thought this was a LitePlacer - it only has one nozzle.  I assume you really mean nozzle tip #1.  Are you calibrating the nozzle tip run-out after changing nozzle tips?  Can you post the log from after nozzle tip calibration? 

[Litterio Andrea Guainella]

Yes Tony, this is a liteplacer. I set to calibration ever on nozzle tip change (for all nozzle tip) with fail homing.

The calibration seems it works.

Today are away from home so I can't post log data but I saved this image for test (I will post trace log file).

I haven't been very lucky with this liteplacer. Yet many here in the group use it and after simple adjustments it works very well.

I can't understand where I'm wrong. Probably the belt are too tight? Or too slow? 

I don't know the right tension they should have but from the videos on youtube it seems that my liteplacer is all ok and instead it has many problems. 

The stapper motors (mostly X) that tears and loses the steps (I try to edit in tinyg and opnp tons of jerk/acceleration value). How to solve?

After a long use (about 30 minutes) I am forced to homing to realign the references as they are all out of phase.

The repeatability tests do not give satisfactory results.

Maybe I mounted the frame wrong? The results of the X calibration are these: 

Results:
Detected Backlash: 0.173mm
Selected Method: OneSidedPositioning
Sneak-up Distance: 0.000mm
Speed Factor: 0.25
Applicable Resolution: 0.0125 mm

instead those of Y are these: 

Results:
Detected Backlash: 0.106mm
Selected Method: DirectionalSneakUp
Sneak-up Distance: 0.226mm
Speed Factor: 0.25
Applicable Resolution: 0.0124 mm

I don't really know what to do anymore. I replaced the stock cameras with the ELP (6mm) and the user experience has improved a lot, especially bottom vision. 

I replaced the cp40 nozzle holder hoping to fix the N runout but with no satisy results.

Maybe the tinyg is the problem? 

I am really frustrated indeed I admire the spirit of this group that has always supported me to overcome obstacles and so many thanks for all.

I hope solve all problems.

LAG

[Litterio Andrea Guainella]

Hi Tony,

here last log before installed new version of openpnp.

I also attach new log with new openpnp version.

Do you think there is something wrong?

The problem is same

LAG

[mark maker]

Hi LAG,

I hope you keep your spirit up! 💪 😁 👍

I think there are many things that can make a Liteplacer not work. The construction is really very simple and also outdated.

For instance all the gantries are tensioned with these eccentric spacers (3). If you make them too tight the gantry will not slide well, which could be one of many explanations that your steppers stall. But if you make them too lose, the gantry will wobble and vibrate, which might in turn stall the stepper when you go very fast, the vibration creating so much friction.

https://liteplacer.com/the-machine/assembly-instructions/gantry-back-plate-step-2-lower-v-wheels/

This is really very bad low-tech compared to even the cheapest linear rails (I'm allowed to say, I have a Liteplacer too 😅).

But it can be adjusted. You just need patience.

As to belt tensioning: From what I researched at the time, a belt should sound a very low sound when plucked on the free side. Like the lowest key on a piano, not like a guitar string. I read that once you overstretched a belt it is damaged and it needs replacing.

_Mark

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/e4a2728e-b0e5-46b7-8adc-8f697bf044a3n%40googlegroups.com.

[Litterio Andrea Guainella]

Thanks Mark for the support. I feel you close. 😊

I have tried them all and with a good approximation I think I have arrived at an acceptable result.

In these days I have disassembled and reassembled everything.

I'll keep this valuable tip in mind. Now the problems are concerning the vision setting up a "ReferencePushPullFeeder" but for that I will open another thread. 

As mentioned in another post, I started buying the material to build another pnp having gained experience with the liteplacer and with the suggestions of the users. 

Maybe I will also ask for specific advice. 

At the moment I bought:

  - aluminum 4020 bar (1.5m long to be cut)

  - rails and skids for X, Y.

Now the most difficult activities remain: designing the junction supports and housings for the cameras. 

I think I buy the head, but for this I will ask users for advice with a new post

LAG