Wrist slaps
Nowadays it's pretty common to see one
of those black plastic bags with a bit of computer inside.
The label warning about static damage
is well known to most people in industry.
A wrist strap and an earthed conductive
table top, although not too familiar, should always be used when
assembling memory and other sensitive parts into a computer.
Some buildings seem to attract static
electricity in huge amounts, whilst in many other places, it
is rare to ever see its effects. At one place I worked we had
nylon carpets and a fairly warm environment and was shocking!
If you walked more than a few yards then sat down and touched
anything, say when you pulled your metal framed chair towards
the desk, a loud crack and an enormous spark would leap from
your finger. This is the sort of phenomenon one must guard against
when handling printed circuit boards.
Most boards carrying semiconductors
are quite safe from damage.
When an integrated circuit is connected
up to power rails, even when no power is applied, it is normally
pretty safe from damage because most modern CMOS has gate input
protection, using zeners which limit the voltages, seen by the
chip, to safe levels. There are some devices however that cannot
use protective circuitry for fear of introducing errors into
their operation. Precision analogue switches usually fall into
this category, certainly the ones that were around in 1975.
One equipment, we developed, gave us
no end of trouble!
There were about 20 equipments, each
using eight printed circuit boards carrying eight 8-way analogue
multiplexer chips. These devices were very expensive; in those
days each board probably cost us about £500. In 1975 that
was 2 or 3 months wages.
Most of the boards coming from Production
were faulty. Usually all the analogue channels bar one group
would be OK. That is 56 channels on a particular board were OK
and 8 were not. The engineer would tie a label on the board handle,
identifying the faulty chip, return it to its black plastic bag
and send it back to Manufacturing for rework.
When the board came back the faulty
channels had been fixed.
The problem was that the other 56 were
now duff.
The engineer would eventually get round
to testing the board, diagnose 7 faulty chips, tie a label on
the handle etc.
Bill, the Head of Quality Assurance
got involved when it started to look like we would miss our factory
acceptance date and we gave poor reliability of the analogue
switches as the reason.
He soon discovered the fact that boards
were being fixed and coming back worse than when they went for
repair.
The odd thing, he also discovered was
that when a board with 7 faulty chips was fixed it came back
with just one faulty chip.
In fact some boards had been round the
loop loads of times and no-one had really noticed. The pattern
was always the same and entirely predictable. He summed it up:-
(1) A lot of boards from Manufacturing
arrived with one chip faulty.
(2) Boards sent back with one chip faulty
came back with seven faulty.
(3) Boards sent back with seven chips
faulty came back with one chip faulty.
And by now we were having difficulty
buying chips and special shipments were having to be flown over
from the US.
Two engineers accompanied Bill from
QA around Production.
"Let's see the manufacturing documentation",
Bill asked Don, the Production Manager.
"Here you are Don said", pointing
out the instructions about an earthed bench and the use of a
wrist strap.
"Let's see where the boards are
assembled", said Bill.
Sure enough all was in order. He was
shown a nice tidy bench using a large sheet of aluminium, suitably
earthed, and a proper wrist strap.
"What happens to the assembled
boards", asked Bill.
"They're put into black plastic
bags and sent over to be flow soldered was the retort".
We went over to look.
A board was in the process of being
soldered.
One corner of the board came into contact
with the solder bath first and of course it was the corner nearest
to the chip that was usually faulty on new boards. Static built
up on the surface of the solder bath and discharged to the new
board as it neared the hot metal.
NO that was too easy!
The solder bath was well earthed and
later tests showed the problem wasn't there.
Tests after the soldering process revealed
the same fault as predicted at the corner of the board but it
was noticed that some boards were soldered the other way round
but the fault was always at the same corner.
What happened after the soldering process?
The boards were taken over to an ultrasonic
cleaning machine where they were dunked in a solvent to get rid
of flux.
The boards were placed in a batch and
immersed in the fluid and the machine vibrated everything at
a very high rate and the boards came out nice and clean.
Unfortunately they came out knackered!
A simple change to the process was suggested
and we afterwards had 100% working new boards.
What about the rework problem?
We went to see Jack, the chap that removed
the faulty chips.
He showed us his bench.
Not quite as pristine as the production
facility but all the necessary bits were there including his
special isolated soldering iron and wrist strap.
"I know all about the procedure",
he said producing a well thumbed document from his desk drawer.
"Fitting of Static-Sensitive Integrated
Circuits", the title announced, and inside, "When fitting
and soldering a chip
.".
It was all there in black and white
and he assured us he definitely followed the instructions to
the letter.
"I don't get it there must be something
wrong", said Bill.
"Here's a typical board, show us
what you normally do".
Jack removed the board from the black
bag and took it over to a second bench piled high with bits of
units and tools.
Shoving everything aside to make room,
and using a solder wick and a very large old soldering iron he
proceeded to expertly remove the marked chip.
Then he went back to his special bench,
donned his wrist strap, picked up his special soldering iron
and inserted a new chip.
"There you are!", proclaimed
Bill, looking somewhat relieved, "You just destroyed seven
good chips", and went off to slap somebody's wrist and to
re-write the newly entitled procedure. "Fitting and Removal
of Static-Sensitive Integrated Circuits".
We never had any trouble after that.
Post script
I bought an LC7821
for a good quality amplifier yesterday. It's a 30 pin 8 way analogue
multiplexer and it cost only £1.23. Is this reverse inflation?