Re: gEDA-user: pick and place ?plugin?footprints.c? for

[Steve Meier <smeier@xxxxxxxxxxxxxxxxxxx>]

The message that I heard was that a generic solution is needed for the
pick and place machines. That libraries from a number of cad programs
were constructed so that the x,y output from them wasn't ideal or even
suitable for programming the pick and place machines.

For a generic solution the gerber files are more or less a common
denominator. Find a solution for precessing the gerbers and you solve
the problem. However the gerbers for what they are have already lost
large amounts of information about the individual components. I wonder
if a survey of the commercial tools would produce an alternative file
format that is common among the cad packages and where the package
information hasn't already been lost.

Pads has an ascii file format that seems very parsible. Do the other cad
programs support the pads ascii format? If they did then perhaps a
program that generates centerized x,y coordinants from that file format
would be more exact and less human interaction then using gerbers.
Similarily I would love to be able to import/export the pads ascii from
pcb.

Steve Meier


On Tue, 2008-02-26 at 10:24 -0600, John Griessen wrote:
> Dave N6NZ wrote:
> > 
> > Dan McMahill wrote:
> >> Dave N6NZ wrote:
> 
> >> Still, the suggestion about letting a footprint optionally include this 
> >> information to deal with "problem" parts might be an option.
> > 
> > Yes, even as I suggested it I considered it a band-aid.  If PnP info can 
> > be computed reliably, it should be done automatically. But let's face 
> > it, in every footprint library there are some footprints that are 
> > "higher maintenance" than others.  With a manually entered fall-back 
> 
> If a footprint contained another reference mark, (pick ooint), besides origin, it seems to
> me it would help.  Think of SMT connectors with offset solder islands for strength of
> grab to the board.  They're not symmetric, and hard to analyze so a mark that comes with
> them is the only way I can think of.  Since the mark has the single purpose of defining where
> is a flat spot your pick and place machine can grab, it wouldn't move around as people
> made footprint variants.
> 
> =========================
>  > Golden rule,  ask user on export.
>  >
>  > We could even show this data on a layer in PCB, or view this in gerbv
>  > as a ball at the centroid with a stick and arrow in the direction of
>  > the axis.
>  > This would allow a human to verify the axis and rotation.
>  >
>  > For pick and place if we are consistent than that is half of the battle.
>  >
>  > Hardkrash
> 
> [jg] I like this idea.  PCB seems the more likely place for the recognition code.
> The gerbv developers say a gerbv library call exporting to pcb format is "on the horizon", so
> a PCB plugin might call gerbv for the starting point gerber data to operate on.
> 
> I still see one area not discussed much, I don't think..), yet needing tricky recognition:  How do
> you identify which pads are in a footprint when starting from gerber data?  Especially
> if some libraries of footprints leave off silk, such as 0402s 0201s etc...?
> That step seems to require looking at pads plus trace exiting pad to decide
> which pairs of teeny pads go together for 2 terminal parts in rows.
> 
> Dave N6NZ wrote:
> ====================
>  > And then there are BGA's..... I'm not going to think about them tonight.
>  >
> 
> Those and land-grid-arrays need two fiducials near them says lilbro.  They aren't the big time-burners.
> Probably a pick mark is hard for non-symmetric land-grid parts, but the fiducials save you anyway.
> Probably OK to leave as "user supplied marks" and worry only about the rest of the part types.
> 
> 
> John Griessen
> back from Kinsale 7 March



_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user