[log in to unmask] said:
>Interesting discussion on using rubber erasers to "polish" fiducials.....
>
>I presently have a customer who has to polish his fiducials in a similar
>manner before his vision system can locate our boards. I want to try and
>help this customer get out of this practice. The board has a HASL finish.
>
>Can anyone shed any light as to the root cause of this issue and reccomend
>solutions (preferrably board related fixes - if there are any).
The vision system is programmed to look within a prescribed area for a
distinct shape such as a circle or diamond, and to calculate a centroid
location. The pattern seen is affected by how much light reflects, and HASL
finished boards vary greatly in reflectance, from dull to mirror bright.
The amount of light seen in the common circle pattern gives the "digital"
appearance of an oval, a lumpy cookie, or a pile of mashed potatoes.
You could alter the software to be more forgiving about what it considers
to be a circle, but if this is done, the risk is a centroid result several
mils off true location, and the pick and place machine would blithely place
all the parts with an erroneous shift. IMO, most systems are "conservative"
in accepting the programmmed shape, and simply reject the fiducial mark if
it does not find a regular pattern.
This can all be greatly affected by the incident lighting angle or the
board position. Our operators here quite often simply hit the "re-try"
button, and it will work, or move the board a hair in the conveyor, and
re-try.
The old eraser trick will do the job, too.
My best suggestion is for you to reduce the HASL coating thickness to a
minimum, and monitor the tin-lead ratio as well as contaminats like Cu.
Eutectic solder is usually the "brightest", but if you shift off that peak
much, it will quickly turn greyish, as the metallographic effects of lead
or tin-rich regions makes a big difference. Perhaps (and I certainly don't
know) it would even be better to maintain some tin content somehwat "off"
the eutectic - so long as it is consistent.
I'd bet there has never been one day here we have gotten 100% recognition
of all fiducials going through our machines - and in most cases I consider
the boards to be within normal variation regarding the HASL coating. I'll
venture to say that when the problem increases, the HASL thickness
correlates pretty well.
> I have also read in the forum that gray scale vision systems can remedy this.
Our ICOS M1000 greyscale system (circa 1990) is imperfect in its
recognition, it will mostly reject a fiducial altogether. Recently I saw a
brand new P&P system (using greyscale) being demo'ed which rejected 1 mm
HASL finished circles irregularly.
This problem could doubtlessly be reduced by improvement in the lighting,
or by some method of automatically re-trying the recognition from a
slightly altered angle, but I am not aware of any machines which do this.
The problem results from the tremendous variation in reflectance of the
smooth (or perhaps not so smooth) metal which is three-dimensionally
variable (look under a microscope at the fiducial mark - it may appear to
have a surface like the moon, or like a drop of oil on water). IMO, this is
a potential advantage for the OSP finish, but our machine had to be
re-programmed to recognize marks when we changed one part from HASL to OSP.
Reminds me of the arguments I used to be dragged into about whether solder
joints were sufficiently "bright". You can have relatively minor changes in
the surface of freezing metal change the appearance radically, as surface
irregularites at the submicron feature size will be easily noticed, either
by our eyes or those of a machine.
regards,
Jerry Cupples
Interphase Corporation
Dallas, TX USA
http://www.iphase.com/
***************************************************************************
* TechNet mail list is provided as a service by IPC using SmartList v3.05 *
***************************************************************************
* To unsubscribe from this list at any time, send a message to: *
* [log in to unmask] with <subject: unsubscribe> and no text. *
***************************************************************************
|
