I have a problem with vision alignment for a 6-pin tact button. Each board that I populate has about 75 of these things on them, and the vision alignment doesn't know what to do with them because of the awkward arrangement of pins. I usually need to manually rotate (30 degrees or more) a dozen or so on every board after the job is finished. Is there any way to define a specific comparison image to force alignment? Or to choose an algorithm that works well with this shape? Or do I need to just turn vision off for this part and hope they come out of the component tape straight enough? (I haven't tried this yet.)
This image shows the part and alignment results from an example placement. Brightness/contrast seems okay, but it just won't rotate to the correct alignment; it keeps bouncing around at odd angles. I've tried every alignment preset (you can see #11 is selected here), but the result seems to be the same in all cases.
Thanks for the reply. I tried using the 00:General alignment setting, and I played with the brightness/contrast settings to get it to include the whole body (didn't seem to help since the added inner visual area is smaller than the pin extremities. I also tried disabling angle alignment altogether, which actually had the best performance since the component tape keeps them pretty straight and the part is big enough to allow for a small amount of error in rotation.
However, even with angle alignment disabled, I'm running into what appears to be a bug in the machine's behavior. Even with large nozzles on this switch (I've tried both of the largest I have with the same results), the head rotates *AFTER* the vision system finishes. The part is perfect while over the camera, then rotates about 30 degrees as it moves from the camera to the location on the board. I captured this in slo-mo video so you can see:
1. The part is picked up from the tape by Head 2 2. The part is rotated 180 degrees (correct) as it moves from the tape to Camera 2 3. The part finishes alignment and the light goes off 4. The part rotates about 30 degrees AS IT MOVES from Camera 2 to the PCB location 5. The part is placed at a very wrong angle
(I had to break the link because the forum won't let me post links yet)
I noticed an interesting pattern to the failures, which makes me think it is a bug and not a configuration or vacuum or nozzle selection error. It's not a perfect pattern, but it is a strong correlation. The parts that need to be placed at 0 or 180 degrees are almost always straight, with no post-alignment rotation. But the parts that have 90 or 270 degree placement are almost always very wrong.
In the image, you see 5 rows of switches, from top to bottom:
- ROW 1 @ 0 degrees: Placed by machine, no errors - ROW 2 @ 90 degrees: Placed by machine, all errors but I fixed them mid-job by hand - ROW 3 @ 0 degrees: Placed by machine, no errors - ROW 4 @ 90 degrees: Placed by machine, all errors without manual correction - ROW 5 @ 0 degrees: Placed by machine, one error only
There's no good reason for this that I can think of. All parts came out of the tape the same way, and came off the visual alignment step almost exactly right.
For reference, my machine has a "V2.03" sticker on it, and I'm currently running Vision Placer v2.055. I have not updated firmware (not sure how) nor up to the latest software since I got a relatively stable setup and I figured I might as well not risk breaking something. Should I update anyway? Is the above behavior something known (or suspected) with a fix included in a later release?
Zitat von jrowberg im Beitrag #3I tried using the 00:General alignment setting, and I played with the brightness/contrast settings to get it to include the whole body (didn\'t seem to help since the added inner visual area is smaller than the pin extremities. I also tried disabling angle alignment altogether, which actually had the best performance since the component tape keeps them pretty straight and the part is big enough to allow for a small amount of error in rotation.
I think you misunderunderstood something. Even if you include the whole body all pins need to keep in visual area! Do not update the software now the firmware of your machine is really old and can only get updated by exchanging CPU PCB.
Did you try to pickup part and place without visual alignment? It is possible that it stay not in position under nozzle because of not enough friction. Use a slow rotation speed too!
Zitat von mbruch im Beitrag #4I think you misunderunderstood something. Even if you include the whole body all pins need to keep in visual area!
I kept the visual area big enough to include the entire part, yes; I only adjusted the brightness and contrast settings so that all of the visible surface area of the switch would show up as white, rather than just the pins. I mean that it didn't help because the new area that became white is just a small square inside the virtual square perimeter defined by the six pins, so the alignment algorithm doesn't behave any differently.
Zitat von mbruch im Beitrag #4Do not update the software now the firmware of your machine is really old and can only get updated by exchanging CPU PCB.
Is this something that you would recommend for performance/stability/functionality? If so, how do we make it happen?
Zitat von mbruch im Beitrag #4Did you try to pickup part and place without visual alignment? It is possible that it stay not in position under nozzle because of not enough friction. Use a slow rotation speed too!
I did try that, yes. It doesn't seem to help. I usually run jobs on medium speed, and I tested with low speed without any change in behavior. I have good air pressure and my nozzles are very large (CN200 and CN400 tested), so I don't think the problem is friction. You can see in the YouTube video how the final unexpected rotation occurs gradually as the head moves from Camera 2 to the PCB. The pattern of straight vs. crooked based on the target part angle (0/180 vs. 90/270) seems too predictable to be a friction problem.
Is there any way log or otherwise capture machine commands for debug purposes to help narrow down possible causes?
Is the nozzle really rotating? Would it be possible to shoot the video again and stick a piece of white tape on the nozzle to see if the nozzle is rotating?
In new software you can use the log function to write commands in log file. The white status window show all commands in UI this content is written to log file.