"John Larkin" wrote in message
...
On Sat, 05 May 2007 15:51:59 GMT, "Jeff L"
wrote:
"John Larkin" wrote in
message
.. .
This is the Lantronix Xport module.
http://www.lantronix.com/device-netw...rvers/xport.ht
m
l
This particuler unit randomly, a couple of times a week maybe, draws a
lot of power and gets very hot, 80C or so, and still works, but
flakily. So we decided to rip it open and see what's inside.
I love those LEDs.
At least one looks to be bi - color as it looks like there are two bond
wires and two LED dies in it.
There seems to be significant manual assembly needed for these, as it
would
be hard for a robot /selective solder to solder many of the through hole
components (LEDs, inline header, ground wire, connector going to the
outside
for your PCB, RJ45 pins (they are SMT on both sides of the board), the
transformer wires (these may be wave or selective soldered)).
Yeah. The board-board pins are stiff, old cordwood style. Must be hell
to rework.
Hot air will make rework simple, as well as a really wide tip that can touch
all pins at once.
The manually
assembled parts look to have significant weird looking flux residue
around
them, at least on the blue board.
Ugly yellow no-clean flux everywhere.
I'm surprised they did not clean them - this product is generally meant for
low volume specialized equipment that is intended to have a fairly long
lifetime. I generally encourage people to wash the boards if the product is
intended to last more then 5 years, and especially if it will not be used in
an reasonably controlled environment with reasonable humidity.
It's a different story when your building mass produced consumer stuff
(junk?) where profit is low and unfortunately the product life time is
measured in months.
Weird ground wire (I assume) going to the rear case.
Right.
I am quite surprised they used a 1.6 mm thick PCB, and one close to it. I
suppose they have the room (there is surprisingly a lot of room left
over).
The simple top PCB is at least a 4 layer PCB.
It's weird they used a custom set of pulse transformers and potted them
to
the PCB. They must have had problems finding one small enough, but then
again the ones in PCMCIA cards were quite small.
I'm surprised the Atmel memory IC is a BGA - all the ones I've currently
used and seen were CASON 8 or oversized body SO8's
Balls on the top ic are directly opposite balls on the other one.
Ballsy, somebody called it.
That is a bit "Ballsy". The BGA on the bottom of a double sided reflow (this
is when one side of a PCB is built completely, and then the other side is
built - as in the case of about 99% of automated double sided PCB assembly)
will have it's balls stretched out as the solder ball surface tension holds
it from falling.
I don't know how that effects reliability as we don't encourage this type of
design, but a demo board from a large semiconductor manufacture had a BGA
that was on the bottom of a double sided reflow. The BGA had a flaky ball
and had to be brought up to reflow temperature on a rework station in the
upright orientation to fix it.
Placing BGA's opposite from one other could produce some interesting routing
and via issues, but if blind vias are used (which is likely, since I don't
see many vias, especially the ends of ones going to internal planes only),
that would be eliminated. The reflow profile could get a little tricky, but
the parts are small in this application, so I doubt that would be an area of
concern.
There are a bunch of 0402's. The pads width and overall length looks
good,
but the gap between the two pads looks a little large - possible
tombstone
or worse, drawbridge problems, but without having the actual dimensions
or
looking at the board in more detail, it's hard to say.
There is no evidence of something getting hot - usually the PCB will
darken
in high heat areas, and if really hot, the solder joints will melt a bit,
giving a cold solder / wrinkly look to them
Every once in a while, Icc would jump up and the case would go up to
80C or so. It still worked, mostly. It would take a power cycle to
reset it. Sure sounds like cmos latchup to me.
That sounds like latchup also - Maybe a flaky BGA ball caused it ;-)
Standard factory
reaction, "that's never happened before."
John