[uCsimm] finepitch SMD soldering

From: Alex Perry (alex.perry@qm.com)
Date: Wed Jan 19 2000 - 14:27:49 EST

> > > > A friend of mine has a rather novel approch to these
> > > > parts, first he solders them to the board (without
> regard to solder bridges), then
> > > > he uses solder wick to pick up the exess and removes
> any bridges!
> > >
> > > This is the best way if you don't have a hot air tool.
> It works just fine
> > > with parts of _any_ pitch (including the TQFP100 parts
> used on the uCsimm).

Our preferred way to do these parts is to give them to someone with
the machine placement and the controlled hot air systems that'll
trivially get them right because it's cheaper than paying someone's
salary to do it the painstaking way. However, you need to be doing
several (i.e. not the prototype) before this is a reasonable viewpoint.

Usually, I'm doing logic chips or obvious chips like processors/ram etc.

Take the pads and wet them with a big blob of solder and plenty
of spare flux. This is because the solderplating done by the board
house is not necessarily the same formulation as yours ... and it
may have oxidized enough to not wet very cleanly to your solder.
Don't try to leave solder behind, just make sure it wets nicely.

Put the chip down, very carefully align the pins to the exact middle
of the pads using a microscope or large magnifying lens on a stand,
then use a large blob of solder in opposite corners to hold it in place.
These blobs should be holding down several pins each for strength.

Now put down lots of solder along all the other pins, one side at a time,
until everything looks really nice and definitely no dry joints. If each
side is one large solderbridge, that is fine, for now.

When done, paint cold flux over the big solderbridges, lay a fresh piece
of copper braid over the row of pins and wipe a large iron tip along to
suck up the solder that is spare. The combination of the wick and
sufficient flux should lift all the bridges very quickly. DO NOT TRY
THIS WITH A SOLDERSUCKER ... like someone did who I described
this to ... it puts 'knots' in the hair along the sides of the chip. Nasty.

Plan ahead for testing the chip for solderbridges; visual inspection is
great and is the first step, but figure out how to do more than that.

For example, with in-circuit-programmable logic, make up a file that
makes the whole chip run as a multifrequency divider. I do not mean
as a binary counter, because the high end bits will be running
really SLOWLY. You want to have the clock divided by 1 for
pin 1, divided by 2 for pin 2, etc etc. Then you can use a frequency
counter to very quickly check all the individual pins out.

For example, with a memory chip that is not boot memory, plan
to write a ram tester for the POST and figure out how to have
the code report where the error is back to you.

For example, with a memory chip that is boot memory, you should
be able to watch each of the address pins count through the
memory fetch at different frequencies, and watch each combination
of two data pins on a two channel scope to look for differences.

Once the processor is booted, it can check everything else for you.

Remember, it is quicker and easier to design-for-test ... if you have
a board with components only on one side, you can plan to make
all traces accessible on the other side at least once and have
the board-house perform their golden-board checkout twice ...
once for the bare board and once for the unpowered stuffed board.

This message resent by the ucsimm@uclinux.com list server http://www.uClinux.com/

This archive was generated by hypermail 2b30 : Sun Apr 07 2002 - 00:01:34 EST