Hi Phil and
others,
I agree that having
pin one, and in this case every tenth pin, distinguished by pad shape aids in
debugging. But as a circuit designer, I have to point out that you only
debug the board once but manufacture it forever. Though I have to debug
the board initially, manufacturing's long-term needs are more important than my
short-term ones. As for field service technicians preferring this old
method to make troubleshooting easier, I would point out that most companies no
longer find it economical to have field service people debugging boards to the
chip level. It costs way too much to train and equip field service people
to this level of expertise and my experience is that they often wind up doing
"shotgun" replacement of parts and poor quality rework. This makes the
feedback to engineering as to failure mode quite useless. After a board
has undergone such "field repair", it is often impossible to tell what, if
anything, was ever wrong with the board. After seeing some of these
beauties, I would never let them go anywhere but the trash
bin!
IMHO, it is more
economical to swap out bad boards and send them back to one place that
is equipped to handle fault diagnosis. Besides, if you want to keep tabs
on failure modes and track down root causes to improve your designs and
processes, this is the only way I've seen that's effective. Once you do
this, you often find out that many of the returned boards are not actually
bad. The technician often does more than one thing on site and it turns
out to be something else that was the failure: an unseated or intermittent
connector termination, a software parameter, a different board if more than one
was returned, etc.
With this in
mind, my suggestions is to make all the pads on surface mount parts the
same and denote pin one with a single silkscreen dot (outside the component
body, please). I feel this meets everyone's requirements well enough:
manufacturing can inspect for orientation and the back end repair folks have pin
one marked. For through-hole parts where there is no back side
silkscreen, use a square or rectangular pad on pin one that has the
same length and width as the other pads.
When we designers
and manufacturing engineers do our jobs right, there are very few boards
returned for repair. When you consider the economics of PCB assembly
manufacture, it makes great sense to design for manufacturing and to design
for test, but very little sense to design for repair at the board level.
There are some special exceptions to this where the board cost is extremely high
and the total number manufactured is small, but by and large I believe this is
the way to go.
Regards,
Seth Goodman
Goodman Associates,
LLC
hi,
the
designer is adding an index which helps test, integration, and rework
personel count pins (for example, 'add a jumper between u2-187 and
u18-41). i'm assuming that these ic's have relatively high lead
counts. it's my experience that these features can be extremely
helpful.
phil
Hi
all,
We're having a board lay-out done outside, and as with new
designs, I will get asked to look at the preliminary gerbers to see if I can
see anything out of place that might need changing.
I looked the the
gerbers that were given to me and the first think I noticed was that on some
IC footprints, there was 1 rectangular pad, then 9 pads where the corners
were rounded, then a rectangular pad, and on and on...
So I fed that
back to our engineer here who is over this project, then he fed that back to
the individual who is doing the lay-out. Our engineer here then came back to
me and asked if that would be a problem, and I told him it could be, but why
not make all the pads uniform? Go to: http://www.stevezeva.homestead.com and
look at "OrCad" to see what I mean.
His reply was that the
individual doing the layout told him he would have to go in and manually
edit each one of the pads that were rectangular, and that it would be too
much trouble. He said this is the way OrCad put the footprint down, and if
it was such a problem, why would OrCad have this footprint in the library?
Then he added that the IPC-SM-782 says that you can either use rectangular
pads or optionally round the corners, so that tells him you can use either
one or both, and it doesn't make any difference. So he told the layout
person to leave things as they are.
I don't know anything at all
about OrCad, but something tells me there's an operator problem here. I've
never seen a footprint like this before, and no matter what I say, I can't
convince our engineer here that the footprint needs to be fixed.
HELP!!! PLEASE!!!
Thanks,
-Steve Gregory-