Roy said:
Production management as well as QA should be rode out of town on a
rail for this behavior then.
One must NEVER bend a lead AT the package!
It stressed the lead frame internally in the device, and can lead to
welded lead connection failures internally, as well as package
fracturing.
ALL mechanical stress placed on a lead MUST be captured externally.
There is ALSO a minimum lead bend radius to consider as sharply bent
copper leads tend to "work harden" and micro-crystalline fractures
WILL occur making the lead current carrying capacity less than half
what it was.
So, even the engineer that designed it should be reamed for not
designing this thing such that proper lead forming practices could be
maintained.
One is not even supposed to bend a resistor lead at the body, much
less more advanced components.
The interesting thing is the environmental circumstances you gave.
It sounds like more is going on than the customer has disclosed.
I learned that lesson the hard way. the very hard way. but with an SOIC8.
It was after discovering the effect of Zout on a gatedrive (or in other
words, if I keep driving bigger and bigger IGBTs wiht the same cheap
little circuit, what happens?) late in the development phase of a small
box of tricks (0.5kW - 5kW). My colleague and I had sorted out why the
small ones all worked and the big ones all died, and we designed a new
circuit to fix the problems (although we kept making the old circuit for
the smaller units, half of the product range).
time being short, and us being in a hurry, our layout guy made a silly
mistake that nobody picked up, and swapped the pins on a comparator. Our
GM interfered behind our backs, and changed our very fast turnaround
prototype order of 10PCBs into 1,500 PCBs (this we discovered when they
turned up). Within a day we had the gatedrive up and running, and found
the swapped leads. doh.
With a bit of butchery we got the damned thing going, and it performed
very well. So off on soak test. 2 hours later, kaboom. Ah ****, back to
where we first started.
repeat many times.
After about a week of looking very, very closely at the gatedrive, I'd
concluded it was, in theory, fine. My colleague then managed to show the
comparator output changing state when heated up sufficiently; cool it
down and it changed back. scope measurements confirmed the physical pins
themselves werent changing, but the output was. We attributed this to
bending the leads of the SOIC8 having damaged the bond wire. when it
heated up, the bond wire went open circuit, and BANG.
some units lasted longer than others; depending, we presumed, on the
degree of damage done to the bond wire, it would survive one or more
thermal cycles before expiring.
We swapped the two pins, did another super-fast re-spin, only bought 10
pcbs this time, and it worked beautifully, first time. So we built a few
hundred thousand of them.
To add insult to injury, the GM attacked us at a budget meeting a few
months later, for "wasting so much money on prototype PCBs". Our
spineless leader didnt have the balls to point out the GM was the idiot
that ordered the extra 1,490 PCBs.
Cheers
Terry
