Dennis Saputelli wrote:

> so how does a 'flying probe' test really work?
> i understand the general idea of a couple of probes walking around the
> comparing connectivity to a 'netlist' made from the gerbers
>
> but it seems to me and i think i read somewhere that this is better at
> finding opens than shorts

It should ALWAYS find 100% of opens, as it should either test all
possible combinations of points on a net, or walk down the net, checking
from one end down to the farthest end.

Checking all possible shorts, especially when opens may be present on the board,
is not possible, due to combinatorial explosion.  So, they have to use some
sort of algorithm to figure out which nets are most likely to be shorted
to another.  nets which pass close to other nets, or have pads adjacent to
another net, are the most likely.  It is SUPPOSED to be correct practice
to run the board again after fixing opens, so that the short detection can
have a better chance of finding a short.

>
> whereas the good old bed of nails would (or could?) find both

I think a flying probe may do better on large boards.  The number of
pins needed for a good-sized modern SMT board can run to the many
thousands.  The cost of the dedicated test fixtures, and the cost of wiring them
up is a killer for low-volume boards.  I suspect many outfits cheat on the
test fixtures, and only place pins on the ends of nets.  If you have a long,
meandering net on an SMT board, it may not go through the pads much,
so many small gaps in the traces could be missed.  If you examine a tested
board closely, you can actually see the marks made by the probes (either
kind) and see how many points are being tested, and whether they are only testing
from via to via, or pad to pad, etc.

When volume gets above several tens of units, then the bed of nails is
needed, as test time on the flying probe machine will become excessive.

> anybody know the down and dirty secrets of all this?

This is one of those dark areas, where the fabricators don't care to have
the buyers know exactly what they are doing.  One reason is that many
fabricators who CLAIM to have in house test, DON'T!  There are test
outfits that will test boards on a few hours notice, and the fabricators
ship stuff all around to whoever has available time on their machines.

I also would not be surprised, since I caught a manufacturer on this, once,
that instead of generating a net list from the gerber and drill info, or
from info supplied by the designer, they test the boards AGAINST EACH
OTHER.  IF they all have the same connectivity, they ALL PASS!
I had a 6-layer board with some very convoluted split power planes in
them.  The fabricator increased the clearance around the non-connecting
through holes on the inner planes without my approval, although I had
already provided the clearance called out on their design rules.  This split
one of the planes into several sections.  All the boards passed, because they
all were split like that.  They couldn't explain how this could happen, but
I had a pretty good idea.

This would be fine if they were supplied a "golden board" from a previous
run, but testing a run against all members of that run allows all sorts of
gremlins to get in, like corrupted Gerber files, Gerber files that trigger
different interpretations of the way to draw something, missing layers,
etc.

Jon

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Forum Guidelines Rules:
* http://www.techservinc.com/protelusers/forumrules.html
*
* Browse or Search previous postings:
* http://www.mail-archive.com/proteledaforum@techservinc.com
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to