I once worked for Burr-Brown in the function modules group. We made the weird stuff: square and square root, vector adders, log amplifiers, and the like.
Mostly I worked on computing RMS modules. These designs were all based on monolithic pairs of transistors, and the same were used as the input stage of operational amplifiers. They were made by gluing a pair of transistors in a ceramic cup and bonding gold wires out to six leads. Then a dot of very hard epoxy protected the bonding wires. We sorted out the incoming parts for the ones with the best-matched characteristics for multipliers, so the function modules got the newest parts.
One day a line tech came by and gave me a box of expensive multipliers that had failed after potting. Normal loss was a percent or so -- this time about 40 percent failed. I don't remember exactly why I suspected the dual transistors, but after grinding down half a dozen modules, I found that the leads were open on one or more of the dual transistors. Snapping off the potting plastic let me read the date code on the edge of the cups. Three date codes had failures.
Eventually this was traced to a materials mixup where the hard epoxy had been replaced by a soft epoxy that moved under the stress of potting -- enough to shear off the gold leads.
It never occurred to me that I should ask before shutting down production to get the bad parts out of the flow. Months later, at a performance review, I found this was a strike against me.
Not long after that review, I moved on to a job at the long-defunct Arizona Gear and Manufacturing Co. There I was dismissed for refusing to certify thermocouple isolation amplifiers for 250,000 MTBF. There were parts in the modules stressed to where they would be lucky to last 2,000 hours (400mW on a 250mW resistor as I remember).
The timing was wrong (1972) for these modules to have contributed to the Fermi-1 reactor meltdown (1966). However, the failure of similar thermocouples on the top of the reactor core made it hard to see the blocking effects on the sodium coolant of a loose plate of zirconium.
This entry was submitted by Keith Henson and edited by Rob Spiegel.
Keith Henson is electrical engineer, a proto-transhumanist, and a writer on life extension, cryonics, memetics, and evolutionary psychology. He has published a Web-book, Standard Gauge, which takes the reader on an exploratory journey into the post-singularity, near-future technology of an artificial intelligence directed clinic in Africa.
Tell us your experience in solving a knotty engineering problem. Send stories to Rob Spiegel for Sherlock Ohms.