Dave N6NZ wrote: > Dan McMahill wrote: > >> This doesn't address your bigger picture problem, but the xy data from >> pcb actually does not depend on where the center is drawn in the >> library. It is calculated as the centroid of the pin/pad centers. So >> for any symmetric part, I can actually say with some confidence that the >> center will be correct in pcb's output. > > Very cool. > >> The orientation is based on an >> algorithm that looks for which quadrant pin #1 is in and then has some >> rules for what you do if it is along an axis. > > I'm a little fuzzy on this. Can you be more specific about how you > calculate rotation?
the description follows http://pcb.sourceforge.net/pcb-20080202/pcb.html#Centroid%20File%20Format the full details start on line 322 of: http://pcb.cvs.sourceforge.net/pcb/pcb/src/hid/bom/bom.c?revision=1.13&view=markup hopefully the code there isn't too bad to read even if you're not up on pcb code internals. The astute code reader will note that only multiples of 90 degrees work and that if you have multiple pins/pads with the same number, they are counted individually for the purposes of finding the centroid. There are probably other ways of confusing the code. I'd welcome someone to poke at the algorithm with a sharp pointy stick and see where it fails. I'm not sure how to make the algorithm properly deal with angles other than multiples of 90 degrees. That is a pretty major assumption. I think before I entertain ideas of modifying that code I'd like to get a regression testsuite in place. I'm thinking of several individual footprint files that are each exported from the command line and the output compared to a hand verified output. I suppose one large layout might be ok for this too since you could sort the output on refdes so that a diff took you right to which element got changed. That would probably run faster. -Dan _______________________________________________ geda-user mailing list [email protected] http://www.seul.org/cgi-bin/mailman/listinfo/geda-user
