Problem compositing TQFP-100

[JW]

Hi all,

I had visual compositing working on a TQFP-100 yesterday, but have made some changes to the machine since (up camera focus Z height, PCB heights etc, nothing major).

Running a job with a TQFP-100 in, paused.

This is the error shown for the package.

I understand for compositing to work, the package must have some symmetry, in either X or Y, which the TQFP-100 does. Not to mention it was working OK yesterday.

I'm not sure I fully understand the error message.

Anybody seen this before on a basic package like this?

[JW]

I should add, visual compositing is still working for other packages on the machine, so it's not some global setting I've made or changed, it seems to be this package only. I've re-generated the footprint associated with this package too, but no change.

[mark maker]

Hi JW,

first, and just to exclude that possibility for sure, please check that the footprint pads are symmetric.

Aymmetry should only ever happen, when you imported them from Eagle or KiCAD, and they are not centered in the E-CAD lib you are using. This should really be fixed on the E-CAD side! Or it could happen, when you hand-modified the footprint. Conversely, when you generated the footprints using the built-in generator, it cannot happen.

Then next check your pick tolerance as set on the nozzle tip you are using. Maybe it is too high.

It's a bit complicated to explain, usually users need not understand this, as everything is automatic. 

For smaller ICs you can usually use an "air-corner", i.e., the corner is, where the mental extrapolation of the pin edges meet. This is also a nice corner because it is both X and Y symmetric on square parts, so we only need two corners to nail them (center and rotation).

Larger ICs have the pin rows farther apart over the IC body corner, so this "air corner" extrapolation becomes too unstable (also depends on available camera view size). So we need to detect the corner directly on the pin row, as shown below. But to detect that corner, it must be isolated from the other row over the IC body corner, also taking into account pick tolerance, i.e., how far it could be shifted when we first see it in the camera. The pick tolerance therefore determines minimum and maximum circular mask sizes (indicated by the yellow circles), i.e., whether you can isolate corners from each other. If pick tolerance is too large, the other pin row will come inside the mask and spoil detection. Vision Compositing predicts that and as a consequence no eligible corner is found:

Also note that you need three instead of only two corners, because they are no longer in a square formation.

_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/ede7bf33-69c0-4d74-98b5-27f00b5d7b5an%40googlegroups.com.

[JW]

Thanks Mark,

That all makes sense, I think actually I did increase the pick tolerance on this part on the assumption that it's picked with a ~8mm nozzle on a 14mm body, so in theory was OK with large pick tolerance as long as it was corrected by up vision. I'll look into this in a couple of hours, this'll probably be it!