Issues with part rotation with autoFeeder/bambooFeeder mounted on the back side (Y max side)

[Antal Levente]

Hello all, 

I'm having hard times with setting up my MicroSMT pnp machine with feeders installed on the back side (Y max side). 

The problem appears at the part rotation, I can't get it working properly. Please have a look at the videos below:

Custom panel: 

 - https://youtu.be/oLudEceP4_U

Demo board:

- https://youtu.be/vtT17JdoJyg

- https://youtu.be/hAB6vtR1rJE

OpenPNP version: 

I'm using Bing feeders, configured as BambooFeders. Everything works fine with the sprocket hole detection, the part is picked up correctly,  (the first suspicion was that the holes are on the wrong side. Based on the content from the help popup from the part rotation field the feeder can be placed in any direction) then vision detects the part. In some cases the parts are placed with 90 degrees of face to the expected angle. I've tried the following values in the part rotation field: -270, -180, -90, 0, 90, 180, 270 without success. The most confusing thing is that only some components are going to have incorrect rotations. When the feeder is mounted on the Y0 side(Front), everything works as expected..

Is there a rule that back side (Ymax) should be avoided or could be just a wrong setting on my side or maybe a bug?

I've attached my config xmls.

Any advice is welcome. 

PS: I'm new in P&P area, this is the first PNP machine that I've seen live and the aim is to replace my staying time in front of the scope - any lammer issue/missconfig can happen - also my top camera is off a bit, will calibreate and fix lens movement once the new mounts arriving :)

Regards, 

Levi

[tonyl...@gmail.com]

Is this a random thing or does the same placement (with the part picked from the same feeder) always get placed incorrectly?

[Antal Levente]

I eas testing with a single feeder, same  placement each time, is not random, it changes the pattern when part rotation is changed but there are always components placed incorrectly (same components without setting change). 

--
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/2c931b3c-b9a7-4076-870f-ecfb7e4c8976n%40googlegroups.com.

[tonyl...@gmail.com]

First question, have you verified that your nozzles really rotate by the correct amount and in the correct direction? I know that's very basic, but it wouldn't be the first time someone had their controller misconfigured. Just put a mark on the side of the nozzle and verify it moves correctly when using the jog buttons. Note that positive angles are counter-clockwise when looking down on the machine from above.

I looked a bit at the code for the Bamboo feeder and didn't see anything obvious that would say it shouldn't work when installed on the back side of a machine.

Assuming you've followed the standard convention for two terminal devices where at zero-degree rotation, pin 1 is to the left and pin 2 is to the right, the correct rotation in tape should be either +90 or -90 degrees (+90 puts pin 1 away from the tape holes and -90 puts pin one towards the tape holes). So, for now use +90 degrees. Now, what happens when you select the feeder and click the "Move camera to the selected feeder's current pick location" button at the top of the feeders table? The crosshairs should be centered over the pick location with cyan crosshair pointing in the positive Y direction (up in the camera view).

The other thing that would help is for you to supply a log file (with the logging level at TRACE) of running a job with just two placements - one that works and one that doesn't.

[Antal Levente]

Hello,

Rotation seems to be ok: https://youtu.be/c76r-XaAzCY

During the last tests i've seen some strange detection results when the back-feeder was used - see pictures below. As far as I understand, the front always should be on the long edge of the component. When  i'm using the front-feeder - all good, everything is detected correctly at each pick. When  i'm using one from the back side these issues happen each time. 

I'm also attaching the trace log. 

Also uploaded a longer video with 4 runs for the demo board - populating only the 0805 components in the circle from the top layer. first 2 runs are done from the back feeder - the last 2 from the front feeder: https://youtu.be/SoOGZCqhOzo

Levi

To view this discussion visit https://groups.google.com/d/msgid/openpnp/35f58cbf-a011-471b-9953-7e53171eab69n%40googlegroups.com.

[Toby Dickenson]

This is 1:33 in your video, as the part first arrives at bottom vision. It looks like a mispick has left the part at nearly 45° on the nozzle, so it is a lottery whether openpnp will correct it clockwise or anticlockwise.

To make this less confusing, I recommend:

1. Add the pad dimensions to the footprint of your packages. This will draw a yellow outline of the part as an overlay on the camera view so that you know where openpnp thinks the part should be.

2. If you update to the latest test branch build (the current beta release; only a couple of weeks from final release) then it will show that yellow overlay on the bottom camera too.

3. Enable size checking in bottom vision, then openpnp will report an error on this type of mispick.

Toby

To view this discussion visit https://groups.google.com/d/msgid/openpnp/CAO-K_z_6VAq1tTeUVgbAAx%3DD72mhTPJ-ETYQtbD5NSbux-A2HQ%40mail.gmail.com.

[Antal Levente]

Hello Toby, 

Actually it wasn't a mispick but I had the pre-rotate value on "Default"  in the vision settings for this package - so the part arrived already pre-rotated to 45 grades for the vision check - and all the issues started from here apparently. Looks like the vision correction doesn't care about this setting and expects the part in its original angle. I turned off pre-rotate (AlwaysOff), turned on size check and each placement is correct - rotation is done after the vision correction.  

Thank you for your help Toby!

Regards, 

Levi.

To view this discussion visit https://groups.google.com/d/msgid/openpnp/CAH35urespnpAiYP8BN2Gy-qN6URWK609-QfTKb5h3FvA0MJjvg%40mail.gmail.com.

[Toby Dickenson]

I'm pleased it is working for you. However, unless I'm missing something, that is not how it should work.

The 45° error I was referring to in that video is between the part orientation and the light blue axis. That axis should mark 0° orientation of the part. The blue axis and part should always be consistent.

What does that yellow package overlay show?

To view this discussion visit https://groups.google.com/d/msgid/openpnp/CAO-K_z83NeYb8UwWt4vbsCMd8eHeCSF%3DonkRh7TjX_W6n9H%3D0g%40mail.gmail.com.

[Antal Levente]

Hello Toby, 

I used my existing installation first then installed the beta version right now. The behavior is the same, with pre-rotate off - it works, if I set Default or AlwaysOn - when the part needs to be rotated I get the error because the detected sizes are out of the configured limits because the part arrives already in the prerotated angle. If the expected behavior is to take in consideration the pre-rotation flags and adapt the vision detection to this - then you pinpointed a bug apparently... 

Here is the video from the test process: https://youtu.be/5fUnye206wg (needs a couple of minutes to be processed by YT).

Regards, 

Levi

To view this discussion visit https://groups.google.com/d/msgid/openpnp/CAH35urdWVhD0e_NgnuU-GDWBFPUfXPkeawaKvSp%2ByZsrVvfzLw%40mail.gmail.com.

[Antal Levente]

One more observation as a possible test result once we are talking about a beta version - after installing the test version, the settings from these dropdowns were empty for all bamboo feeders (it was ok for the rest ) - I had to re-select the values and save..