>> You should use a manufacturer that requests test procedures and will test
>> the cards for you too, as lost of the faults to be found will be process
>> orientated.
>
>hmm.. the economics of testing and who pays for rejects seems intimately
>related. Obviously I don't want an assembler with the attitude
>"we don't know - we put everything in the right place, if it don't work
it's your problem"
>but I wonder just how much testing I can justify on a little 3-4 chip
board,
>especially since it's not likely to fit easily into digital test equipment,
>since it's a half analog/half digital board.

Annie,

Let the economics of the situation determine how much testing you do.  To do
this, try to create a business case with the information:

Calculate what is the expected defect rate from mechanical (soldering) and
electrical defects.  Both these parameters are available from the card
stuffer and the chip manufacturer.  If they aren't, then source your product
elsewhere.

Solder defects will probably be measured in terms per joint and Time-Zero
Electrical Defects will be in terms per chip.  Summing these values with the
part and pin counts on the board will give you an expected defect rate.

Find out what your customers tolerable Zero-Time defect rate is.  Ideally,
your committed to quality levels from your suppliers will sum up to a value
which is less than what your customer expects.  If this is the case, then no
testing is required...

BUT, before going ahead and making this decision, you have to have
understand what your supplier's guarantee is if they miss their committed
quality levels.  Chances are, you will get this IFF:

1.  You get agreement that the circuits do not exceed any manufacturer
specifications.
2.  The card manufacturer agrees that the Card Physical Design will work
efficiently on their equipment and all parts and processes are qualified
(including repair).

This guarantee must be in writing (handshakes and verbal assurances aren't
worth much on a Friday afternoon when the sky is falling) and include
replacement of the finished product, with shipping to the final customer for
replacement.  This is a powerful inducement for them to get it right -
replacing even one card with this type of arrangement can cost them $100 or
more.

If you can't get this guarantee, then you will have to do some kind of
testing.

If you are looking at high volume (10K Units per month or more), then it
would make a lot of sense for you to have an InCircuit Test (ICT) fixture
built.  Modern ICT fixtures can handle both Analog and Digital I/O and could
probably provide you with a functional test at the same time.  For a board
like yours, you're probably looking at 30 seconds or less of "tool time"
(time spent on the tester and is being used with an operator - you have to
pay for both).

To minimize fixturing costs and tool time, you would probably want a
"multiple up" fixture (ie one that can test four or more cards in parallel,
using the most possible test pins in the machine).

The advantage of the ICT Tester is that the manufacturer has all the
documentation to fix problems kicked out by the machine.

For lower volumes, you could probably design a small functional test
fixture.  This can be done for extremely modest prices ($500 or less) and
can run in 30 seconds or less.  With good documentation, testing could be
passed to the board manufacturer.

To get an idea about what I mean by good, take a look at my "El Cheapo"
programmer at:

http://www.rentron.com

with associated "El Debug" application.  The "El Debug" helps the builder
find and correct problems in their design.

Admittedly, your design is probably a lot more complex, but spending time
coming up with a good debug process will help the card stuffer build a
better product, reduce his costs (which are ultimately yours) and increase
your customer's satisfaction.


Sorry for the (not so) short paper on quality and contract manufacturing,
but maybe that will help you make the right decision for your product.

myke