<snip>
> I thought smd-coordinate.ulp gave you this on a silver platter?
This seems to give you X and Y coords, but does not tell you the pad
size. We needed the pad size info to determine how much solder paste
to put on each pad. But this wasn't the tricky part; the pick and
place was a lot harder.
>
> >Eagle doesn't do a very good job in
> >the order it makes the pads.
>
> Sort the .smt and .smb output files on one axis.
Not always faster! :-) I've tried all sorts of things, and having our
code report distance traveled. Even then, we found that while one
sort may have less distance traveled, it might actually take longer to
do on the mill than a different sort with a longer distance traveled.
If you Google "Traveling Salesman problem" you will most likely get
overwhelmed with academic studies... Even simple algorithms such as
"the next pad is the one closest to the current pad" are not always
the best. If I were to spend more time on this, I think I would have
my program try several different algorithms, starting at random pads
on the board, and pick the one with the shortest distance traveled.
More of a Monty Carlo (yea, I know it's miss-spelled) type approach.
But I probably won't, because too many interesting projects, not
enough time...
Over-all, this was a very fun project that we threw together in less
time than some people spent arguing over which is the best way to do
such a thing. And it actually came in handy when we had a crash
project developing the battery management system for the Lithium Ion
powered (www.a123systems.com) world's fastest electric motorcycle
(some videos on my web site as well). Due to the extreme time
constraints, we didn't know all the component values until it was too
close to race day to have a contract manufacture assemble the boards
for us. While the current implementation is not a heck of a lot
faster than assembling by hand, it kept us from going crazy putting on
the over 100+ surface mount components on approximately 40 boards we
assembled.
From my experience, if you are going to build 10+ of a board, and have
the time, it makes sense to have a contract manufacture do the
assembly for you. They can have stencils made up, have all the neat
screen printing equipment, faster pick and place, better regulated IR
reflow ovens, etc. But for prototyping just a few boards, or
extremely tight deadlines, the DIY approach can have some merit. I
expect that in the future, I'll use the CNC Solder paste dispensing on
boards with 100+ components (less than that, it will be faster by
hand). As for the pick and place, well, it takes a fair amount of
effort to setup. For just one or two prototype boards, it is usually
easier to stuff by hand. Especially if you build a section, test it,
debug it, move on to the next section. If I have to make between 5
and 10, I'll probably take the time to set up the pick and place. Any
more than 10, I'll most likely contract it out.
I guess the biggest thing I learned from this project is that surface
mount is not as scary as I first thought. I wish I would have just
tried it a lot sooner. You really don't need fancy equipment, even
like the stuff we built. A good pair of tweezers, and if you are over
35, you may need something like "opto-visors" to help inspect your work.
Well, back to the day job, which is the main reason why I didn't post
about this sooner,
- Steven Ciciora
<snip>